Skip to content

Use Claude Code, Codex, VSCode, Pi, and OpenCode (and 6 other harnesses) for free (1.3B+ free tokens) from your terminal, app, IDE, or phone, and now from the browser with native browser sessions (multi-harness + multi-model) like OpenClaw (voice supported + ToS friendly)

Python

feat: propagate model token limits (#1605)

Ali Khokhard9bac1f7f222
diff --git a/pyproject.toml b/pyproject.toml
index 651e3824..d584a33f 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -4,7 +4,7 @@ build-backend = "hatchling.build"
[project]
name = "free-claude-code"
-version = "5.16.0"
+version = "5.17.0"
description = "Local proxy connecting coding agents to OpenAI-compatible AI providers"
readme = "README.md"
requires-python = ">=3.14.0"
diff --git a/src/free_claude_code/api/model_catalog.py b/src/free_claude_code/api/model_catalog.py
index ecd5b4db..85f4cd5f 100644
--- a/src/free_claude_code/api/model_catalog.py
+++ b/src/free_claude_code/api/model_catalog.py
@@ -53,6 +53,12 @@ class ModelResponse(BaseModel):
input_modalities: tuple[ModelInputModality, ...] | None = Field(
default=None, serialization_alias="inputModalities"
)
+ context_window_tokens: int | None = Field(
+ default=None, serialization_alias="contextWindow"
+ )
+ max_output_tokens: int | None = Field(
+ default=None, serialization_alias="maxCompletionTokens"
+ )
reasoning_efforts: tuple[str, ...] | None = Field(
default=None, serialization_alias="reasoningEfforts"
)
@@ -118,6 +124,8 @@ class _InventoryModel:
provider_model_ref: str
supports_thinking: bool | None
input_modalities: frozenset[ModelInputModality] | None
+ context_window_tokens: int | None
+ max_output_tokens: int | None
def build_models_list_response(
@@ -208,6 +216,8 @@ def _build_direct_models_response(
input_modalities=_serialize_input_modalities(
inventory_model.input_modalities
),
+ context_window_tokens=inventory_model.context_window_tokens,
+ max_output_tokens=inventory_model.max_output_tokens,
supports_reasoning_effort=(
allows_reasoning if view is ModelCatalogView.RESPONSES else None
),
@@ -248,6 +258,12 @@ def _collect_inventory(
input_modalities=(
model_info.input_modalities if model_info is not None else None
),
+ context_window_tokens=(
+ model_info.context_window_tokens if model_info is not None else None
+ ),
+ max_output_tokens=(
+ model_info.max_output_tokens if model_info is not None else None
+ ),
)
)
@@ -260,6 +276,8 @@ def _collect_inventory(
provider_model_ref=model_info.model_id,
supports_thinking=model_info.supports_thinking,
input_modalities=model_info.input_modalities,
+ context_window_tokens=model_info.context_window_tokens,
+ max_output_tokens=model_info.max_output_tokens,
)
)
diff --git a/src/free_claude_code/application/model_metadata.py b/src/free_claude_code/application/model_metadata.py
index edfb2d47..b980bebe 100644
--- a/src/free_claude_code/application/model_metadata.py
+++ b/src/free_claude_code/application/model_metadata.py
@@ -12,6 +12,8 @@ class ProviderModelInfo:
model_id: str
supports_thinking: bool | None = None
input_modalities: frozenset[ModelInputModality] | None = None
+ context_window_tokens: int | None = None
+ max_output_tokens: int | None = None
@dataclass(frozen=True, slots=True)
diff --git a/src/free_claude_code/cli/launchers/aider_config.py b/src/free_claude_code/cli/launchers/aider_config.py
index abf07f9a..9e6ea2b0 100644
--- a/src/free_claude_code/cli/launchers/aider_config.py
+++ b/src/free_claude_code/cli/launchers/aider_config.py
@@ -65,4 +65,8 @@ def _model_metadata(model: ClientModel) -> JsonObject:
}
if model.input_modalities is not None:
metadata["supports_vision"] = ModelInputModality.IMAGE in model.input_modalities
+ if model.context_window_tokens is not None:
+ metadata["max_input_tokens"] = model.context_window_tokens
+ if model.max_output_tokens is not None:
+ metadata["max_output_tokens"] = model.max_output_tokens
return metadata
diff --git a/src/free_claude_code/cli/launchers/cline_config.py b/src/free_claude_code/cli/launchers/cline_config.py
index 95fb9c46..6af0c267 100644
--- a/src/free_claude_code/cli/launchers/cline_config.py
+++ b/src/free_claude_code/cli/launchers/cline_config.py
@@ -113,4 +113,8 @@ def _model_entry(model: ClientModel) -> JsonObject:
"outputModalities": ["text"],
}
)
+ if model.context_window_tokens is not None:
+ entry["contextWindow"] = model.context_window_tokens
+ if model.max_output_tokens is not None:
+ entry["maxTokens"] = model.max_output_tokens
return entry
diff --git a/src/free_claude_code/cli/launchers/codex_model_catalog.py b/src/free_claude_code/cli/launchers/codex_model_catalog.py
index f2954bd3..cb2d6028 100644
--- a/src/free_claude_code/cli/launchers/codex_model_catalog.py
+++ b/src/free_claude_code/cli/launchers/codex_model_catalog.py
@@ -70,6 +70,11 @@ def _codex_catalog_entry(candidate: ClientModel, *, priority: int) -> JsonObject
input_modalities = candidate.input_modalities or frozenset(
{ModelInputModality.TEXT}
)
+ context_window = (
+ candidate.context_window_tokens
+ if candidate.context_window_tokens is not None
+ else 200000
+ )
entry: JsonObject = {
"slug": candidate.wire_slug,
"display_name": candidate.display_name,
@@ -92,8 +97,8 @@ def _codex_catalog_entry(candidate: ClientModel, *, priority: int) -> JsonObject
"truncation_policy": {"mode": "tokens", "limit": 10000},
"supports_parallel_tool_calls": True,
"supports_image_detail_original": True,
- "context_window": 200000,
- "max_context_window": 200000,
+ "context_window": context_window,
+ "max_context_window": context_window,
"effective_context_window_percent": 95,
"experimental_supported_tools": [],
"input_modalities": [
diff --git a/src/free_claude_code/cli/launchers/dsh_config.py b/src/free_claude_code/cli/launchers/dsh_config.py
index 6985b94b..fbde4b85 100644
--- a/src/free_claude_code/cli/launchers/dsh_config.py
+++ b/src/free_claude_code/cli/launchers/dsh_config.py
@@ -121,6 +121,10 @@ def _model_profile(model: ClientModel) -> JsonObject:
for modality in ModelInputModality
if modality in model.input_modalities
]
+ if model.context_window_tokens is not None:
+ profile["contextWindow"] = model.context_window_tokens
+ if model.max_output_tokens is not None:
+ profile["maxTokens"] = model.max_output_tokens
return profile
diff --git a/src/free_claude_code/cli/launchers/hermes_config.py b/src/free_claude_code/cli/launchers/hermes_config.py
index 799e523a..d766ba22 100644
--- a/src/free_claude_code/cli/launchers/hermes_config.py
+++ b/src/free_claude_code/cli/launchers/hermes_config.py
@@ -128,4 +128,8 @@ def _model_override(model: ClientModel) -> JsonObject:
override["supports_reasoning"] = model.supports_reasoning
if model.input_modalities is not None:
override["supports_vision"] = ModelInputModality.IMAGE in model.input_modalities
+ if model.context_window_tokens is not None:
+ override["context_window"] = model.context_window_tokens
+ if model.max_output_tokens is not None:
+ override["max_output_tokens"] = model.max_output_tokens
return override
diff --git a/src/free_claude_code/cli/launchers/model_catalog.py b/src/free_claude_code/cli/launchers/model_catalog.py
index 41a783a4..7ccf71a0 100644
--- a/src/free_claude_code/cli/launchers/model_catalog.py
+++ b/src/free_claude_code/cli/launchers/model_catalog.py
@@ -22,6 +22,8 @@ class ClientModel:
display_name: str
supports_reasoning: bool | None
input_modalities: frozenset[ModelInputModality] | None = None
+ context_window_tokens: int | None = None
+ max_output_tokens: int | None = None
def client_models_from_response(
@@ -108,6 +110,10 @@ def _catalog_candidates(
display_name=_nonempty_string(item.get("display_name")) or model_id,
supports_reasoning=_optional_boolean(item.get("supportsReasoning")),
input_modalities=_input_modalities(item.get("inputModalities")),
+ context_window_tokens=_optional_positive_int(item.get("contextWindow")),
+ max_output_tokens=_optional_positive_int(
+ item.get("maxCompletionTokens")
+ ),
)
)
return candidates
@@ -124,6 +130,10 @@ def _optional_boolean(value: object) -> bool | None:
return value if isinstance(value, bool) else None
+def _optional_positive_int(value: object) -> int | None:
+ return value if type(value) is int and value > 0 else None
+
+
def _input_modalities(value: object) -> frozenset[ModelInputModality] | None:
if not isinstance(value, list) or not value:
return None
diff --git a/src/free_claude_code/cli/launchers/opencode_config.py b/src/free_claude_code/cli/launchers/opencode_config.py
index 332c2255..cf933805 100644
--- a/src/free_claude_code/cli/launchers/opencode_config.py
+++ b/src/free_claude_code/cli/launchers/opencode_config.py
@@ -73,4 +73,16 @@ def _model_config(model: ClientModel) -> JsonObject:
if modality in model.input_modalities
]
}
+ if model.context_window_tokens is not None or model.max_output_tokens is not None:
+ # OpenCode requires both fields and uses zero when either limit is unknown.
+ config["limit"] = {
+ "context": (
+ model.context_window_tokens
+ if model.context_window_tokens is not None
+ else 0
+ ),
+ "output": (
+ model.max_output_tokens if model.max_output_tokens is not None else 0
+ ),
+ }
return config
diff --git a/src/free_claude_code/cli/launchers/pi_extension.ts b/src/free_claude_code/cli/launchers/pi_extension.ts
index 2f648865..1b1467e5 100644
--- a/src/free_claude_code/cli/launchers/pi_extension.ts
+++ b/src/free_claude_code/cli/launchers/pi_extension.ts
@@ -37,6 +37,10 @@ function optionalBoolean(value: unknown): boolean | undefined {
return typeof value === "boolean" ? value : undefined;
}
+function optionalPositiveInteger(value: unknown): number | undefined {
+ return typeof value === "number" && Number.isInteger(value) && value > 0 ? value : undefined;
+}
+
function inputModalities(value: unknown): ("text" | "image")[] | undefined {
if (!Array.isArray(value) || value.length === 0) return undefined;
if (value.some((item) => item !== "text" && item !== "image")) return undefined;
@@ -49,6 +53,8 @@ function modelDefinition(
providerModel: string,
supportsReasoning: boolean | undefined,
input: ("text" | "image")[] | undefined,
+ contextWindow: number | undefined,
+ maxTokens: number | undefined,
): ProviderModelConfig {
return {
id,
@@ -56,8 +62,8 @@ function modelDefinition(
reasoning: supportsReasoning ?? true,
input: input ?? ["text"],
cost: { input: 0, output: 0, cacheRead: 0, cacheWrite: 0 },
- contextWindow: DEFAULT_CONTEXT_WINDOW,
- maxTokens: DEFAULT_MAX_TOKENS,
+ contextWindow: contextWindow ?? DEFAULT_CONTEXT_WINDOW,
+ maxTokens: maxTokens ?? DEFAULT_MAX_TOKENS,
};
}
@@ -79,6 +85,8 @@ export function projectFccModels(payload: unknown): ProviderModelConfig[] {
providerModel,
optionalBoolean(entry.supportsReasoning),
inputModalities(entry.inputModalities),
+ optionalPositiveInteger(entry.contextWindow),
+ optionalPositiveInteger(entry.maxCompletionTokens),
),
);
}
diff --git a/src/free_claude_code/providers/groq/client.py b/src/free_claude_code/providers/groq/client.py
index e48e9766..605e5668 100644
--- a/src/free_claude_code/providers/groq/client.py
+++ b/src/free_claude_code/providers/groq/client.py
@@ -21,6 +21,7 @@ from free_claude_code.providers.openai_chat import (
OpenAIChatProfile,
OpenAIChatProvider,
OpenAIChatRequestPolicy,
+ OpenAIModelListing,
validate_extra_body_does_not_override_reasoning_fields,
)
@@ -49,6 +50,10 @@ _PROFILE = OpenAIChatProfile(
disabled_value="none",
enabled_value="medium",
),
+ model_listing=OpenAIModelListing(
+ context_window_tokens_path=("context_window",),
+ max_output_tokens_path=("max_completion_tokens",),
+ ),
)
_REASONING_FIELD = "reasoning_effort"
diff --git a/src/free_claude_code/providers/mistral/client.py b/src/free_claude_code/providers/mistral/client.py
index 8695ca6b..aaf00cfe 100644
--- a/src/free_claude_code/providers/mistral/client.py
+++ b/src/free_claude_code/providers/mistral/client.py
@@ -40,6 +40,7 @@ _PROFILE = OpenAIChatProfile(
),
(ModelInputModality.IMAGE, ("capabilities", "vision")),
),
+ context_window_tokens_path=("max_context_length",),
),
)
diff --git a/src/free_claude_code/providers/model_listing.py b/src/free_claude_code/providers/model_listing.py
index 852d68c0..76f11b74 100644
--- a/src/free_claude_code/providers/model_listing.py
+++ b/src/free_claude_code/providers/model_listing.py
@@ -1,6 +1,6 @@
"""Provider model-list response parsing helpers."""
-from collections.abc import Iterable, Mapping, Sequence
+from collections.abc import Callable, Iterable, Mapping, Sequence
from dataclasses import replace
from typing import Any, TypeIs
@@ -14,6 +14,7 @@ type RequiredPathValues = tuple[
tuple[tuple[str, ...], tuple[ModelListScalar, ...]], ...
]
type InputModalityBooleanPaths = tuple[tuple[ModelInputModality, tuple[str, ...]], ...]
+type ModelTokenLimitResolver = Callable[[object], int | None]
class ModelListResponseError(ValueError):
@@ -54,6 +55,9 @@ def extract_openai_model_infos(
thinking_sequence_path: tuple[str, ...] | None = None,
fixed_input_modalities: frozenset[ModelInputModality] | None = None,
input_modality_boolean_paths: InputModalityBooleanPaths = (),
+ context_window_tokens_path: tuple[str, ...] | None = None,
+ max_output_tokens_path: tuple[str, ...] | None = None,
+ context_window_tokens_resolver: ModelTokenLimitResolver | None = None,
) -> frozenset[_ProviderModelInfo]:
"""Extract routable IDs from an OpenAI-compatible model-list response."""
model_infos: dict[str, _ProviderModelInfo] = {}
@@ -146,6 +150,12 @@ def extract_openai_model_infos(
fixed=fixed_input_modalities,
boolean_paths=input_modality_boolean_paths,
)
+ context_window_tokens = (
+ context_window_tokens_resolver(item)
+ if context_window_tokens_resolver is not None
+ else _optional_positive_int_at_path(item, context_window_tokens_path)
+ )
+ max_output_tokens = _optional_positive_int_at_path(item, max_output_tokens_path)
if not included:
continue
@@ -154,6 +164,8 @@ def extract_openai_model_infos(
model_id=model_id,
supports_thinking=supports_thinking,
input_modalities=input_modalities,
+ context_window_tokens=context_window_tokens,
+ max_output_tokens=max_output_tokens,
)
model_infos.setdefault(model_id, model_info)
if aliases_field is not None:
@@ -212,6 +224,12 @@ def extract_tool_capable_model_infos(
input_modalities=optional_input_modalities(
_path(item, ("architecture", "input_modalities"))
),
+ context_window_tokens=optional_positive_int(
+ _path(item, ("context_length",))
+ ),
+ max_output_tokens=optional_positive_int(
+ _path(item, ("top_provider", "max_completion_tokens"))
+ ),
)
)
@@ -365,6 +383,37 @@ def optional_input_modalities(
return modalities
+def optional_positive_int(value: object) -> int | None:
+ """Return an exact positive integer, otherwise unknown."""
+ return value if type(value) is int and value > 0 else None
+
+
+def live_provider_context_window_consensus(item: object) -> int | None:
+ """Return the common context limit across every advertised live route."""
+ routes = _path(item, ("providers",))
+ if not _is_sequence(routes):
+ return None
+ live_limits: list[int] = []
+ for route in routes:
+ if _path(route, ("status",)) != "live":
+ continue
+ limit = optional_positive_int(_path(route, ("context_length",)))
+ if limit is None:
+ return None
+ live_limits.append(limit)
+ if not live_limits or len(set(live_limits)) != 1:
+ return None
+ return live_limits[0]
+
+
+def _optional_positive_int_at_path(
+ item: object, path: tuple[str, ...] | None
+) -> int | None:
+ if path is None:
+ return None
+ return optional_positive_int(_path(item, path))
+
+
def _input_modalities(
item: object,
*,
diff --git a/src/free_claude_code/providers/openai_chat/profiles.py b/src/free_claude_code/providers/openai_chat/profiles.py
index 398b2e78..8626d464 100644
--- a/src/free_claude_code/providers/openai_chat/profiles.py
+++ b/src/free_claude_code/providers/openai_chat/profiles.py
@@ -13,7 +13,9 @@ from free_claude_code.core.model_capabilities import ModelInputModality
from free_claude_code.core.reasoning import ReasoningEffort, ReasoningPolicy
from free_claude_code.providers.model_listing import (
InputModalityBooleanPaths,
+ ModelTokenLimitResolver,
RequiredPathValues,
+ live_provider_context_window_consensus,
)
from .base_url import openai_v1_base_url
@@ -103,6 +105,9 @@ class OpenAIModelListing:
thinking_sequence_path: tuple[str, ...] | None = None
fixed_input_modalities: frozenset[ModelInputModality] | None = None
input_modality_boolean_paths: InputModalityBooleanPaths = ()
+ context_window_tokens_path: tuple[str, ...] | None = None
+ max_output_tokens_path: tuple[str, ...] | None = None
+ context_window_tokens_resolver: ModelTokenLimitResolver | None = None
pagination: OpenAIModelPagination | None = None
@@ -280,6 +285,7 @@ OPENAI_CHAT_PROFILES: dict[str, OpenAIChatProfile] = {
path="/models",
collection_field=None,
required_path_values=((("type",), ("chat",)),),
+ context_window_tokens_path=("context_length",),
),
reasoning_delta_field="reasoning",
),
@@ -305,6 +311,7 @@ OPENAI_CHAT_PROFILES: dict[str, OpenAIChatProfile] = {
required_null_field="deprecated",
tags_field="tags",
non_thinking_tag="non-reasoning",
+ context_window_tokens_path=("max_tokens",),
),
),
"siliconflow": OpenAIChatProfile(
@@ -337,6 +344,7 @@ OPENAI_CHAT_PROFILES: dict[str, OpenAIChatProfile] = {
query_params=(("verbose", "true"),),
required_path_values=((("architecture", "modality"), ("text->text",)),),
fixed_input_modalities=_TEXT_INPUT_MODALITIES,
+ context_window_tokens_path=("context_length",),
),
),
"chutes": OpenAIChatProfile(
@@ -356,6 +364,8 @@ OPENAI_CHAT_PROFILES: dict[str, OpenAIChatProfile] = {
exclude_missing_sequence_fields=True,
tags_field="supported_features",
input_modalities_path=("input_modalities",),
+ context_window_tokens_path=("context_length",),
+ max_output_tokens_path=("max_output_length",),
),
),
"featherless": OpenAIChatProfile(
@@ -384,6 +394,8 @@ OPENAI_CHAT_PROFILES: dict[str, OpenAIChatProfile] = {
input_modality_boolean_paths=(
(ModelInputModality.IMAGE, ("features", "image_input")),
),
+ context_window_tokens_path=("context_length",),
+ max_output_tokens_path=("max_completion_tokens",),
pagination=OpenAIModelPagination(),
),
),
@@ -415,6 +427,7 @@ OPENAI_CHAT_PROFILES: dict[str, OpenAIChatProfile] = {
),
thinking_boolean_path=("capabilities", "reasoning"),
input_modalities_path=("input_modalities",),
+ context_window_tokens_path=("context_length",),
),
reasoning_delta_field="reasoning",
reasoning_delta_fallback_field="reasoning_content",
@@ -456,6 +469,7 @@ OPENAI_CHAT_PROFILES: dict[str, OpenAIChatProfile] = {
),
(ModelInputModality.IMAGE, ("capabilities", "vision")),
),
+ context_window_tokens_path=("max_context_length",),
),
),
"vercel": OpenAIChatProfile(
@@ -469,6 +483,8 @@ OPENAI_CHAT_PROFILES: dict[str, OpenAIChatProfile] = {
model_listing=OpenAIModelListing(
input_modalities_path=("modalities", "input"),
thinking_sequence_path=("supported_parameters",),
+ context_window_tokens_path=("context_window",),
+ max_output_tokens_path=("max_tokens",),
),
),
"bedrock": OpenAIChatProfile(
@@ -486,6 +502,7 @@ OPENAI_CHAT_PROFILES: dict[str, OpenAIChatProfile] = {
NO_REASONING,
model_listing=OpenAIModelListing(
input_modalities_path=("architecture", "input_modalities"),
+ context_window_tokens_resolver=live_provider_context_window_consensus,
),
),
"cohere": OpenAIChatProfile(
@@ -675,6 +692,7 @@ OPENAI_CHAT_PROFILES: dict[str, OpenAIChatProfile] = {
),
thinking_boolean_path=("reasoning",),
input_modalities_path=("modalities", "input"),
+ context_window_tokens_path=("context_window", "tokens"),
),
),
"ollama_cloud": OpenAIChatProfile(
diff --git a/src/free_claude_code/providers/openai_chat/provider.py b/src/free_claude_code/providers/openai_chat/provider.py
index 376e8c36..481ce5e8 100644
--- a/src/free_claude_code/providers/openai_chat/provider.py
+++ b/src/free_claude_code/providers/openai_chat/provider.py
@@ -4,7 +4,7 @@ import asyncio
import sys
import uuid
from collections.abc import AsyncIterator, Awaitable, Callable, Iterator, Mapping
-from dataclasses import dataclass
+from dataclasses import dataclass, replace
from enum import StrEnum
from typing import Any
@@ -553,12 +553,23 @@ class OpenAIChatProvider(BaseProvider):
thinking_sequence_path=listing.thinking_sequence_path,
fixed_input_modalities=listing.fixed_input_modalities,
input_modality_boolean_paths=listing.input_modality_boolean_paths,
+ context_window_tokens_path=listing.context_window_tokens_path,
+ max_output_tokens_path=listing.max_output_tokens_path,
+ context_window_tokens_resolver=listing.context_window_tokens_resolver,
)
model_infos_by_id = {
model_info.model_id: model_info for model_info in live_model_infos
}
for model_info in model_infos_from_ids(listing.additional_model_ids):
- model_infos_by_id.setdefault(model_info.model_id, model_info)
+ existing = model_infos_by_id.get(model_info.model_id)
+ if existing is None:
+ model_infos_by_id[model_info.model_id] = model_info
+ continue
+ model_infos_by_id[model_info.model_id] = replace(
+ existing,
+ context_window_tokens=None,
+ max_output_tokens=None,
+ )
return frozenset(model_infos_by_id.values())
async def _list_models_payload(self) -> Any:
diff --git a/src/free_claude_code/providers/openai_codex/provider.py b/src/free_claude_code/providers/openai_codex/provider.py
index 8e18c93c..83d9c505 100644
--- a/src/free_claude_code/providers/openai_codex/provider.py
+++ b/src/free_claude_code/providers/openai_codex/provider.py
@@ -48,7 +48,10 @@ from free_claude_code.providers.failure_policy import (
is_retryable_stream_error,
)
from free_claude_code.providers.http import ProviderAttemptScope, maybe_await_aclose
-from free_claude_code.providers.model_listing import optional_input_modalities
+from free_claude_code.providers.model_listing import (
+ optional_input_modalities,
+ optional_positive_int,
+)
from free_claude_code.providers.openai_responses.presentation import (
MessagesResponsesPresenter,
NativeResponsesPresenter,
@@ -609,6 +612,9 @@ def _model_infos(payload: Any) -> frozenset[ProviderModelInfo]:
input_modalities=optional_input_modalities(
model.get("input_modalities")
),
+ context_window_tokens=optional_positive_int(
+ model.get("context_window")
+ ),
)
)
if not infos:
diff --git a/src/free_claude_code/providers/opencode/catalog.py b/src/free_claude_code/providers/opencode/catalog.py
index 037236dc..38cfcc2f 100644
--- a/src/free_claude_code/providers/opencode/catalog.py
+++ b/src/free_claude_code/providers/opencode/catalog.py
@@ -19,6 +19,7 @@ from free_claude_code.providers.failure_policy import classify_provider_failure
from free_claude_code.providers.model_listing import (
ModelListResponseError,
optional_input_modalities,
+ optional_positive_int,
)
OPENCODE_CATALOG_URL = "https://models.opencode.ai/api.json"
@@ -44,6 +45,8 @@ class OpenCodeModelRoute:
transport: OpenCodeUpstreamTransport
supports_thinking: bool | None
input_modalities: frozenset[ModelInputModality] | None
+ context_window_tokens: int | None
+ max_output_tokens: int | None
@dataclass(frozen=True, slots=True)
@@ -127,6 +130,17 @@ def parse_open_code_catalog(
if isinstance(raw_modalities, Mapping)
else None
)
+ limit = model.get("limit")
+ context_window_tokens = (
+ optional_positive_int(limit.get("context"))
+ if isinstance(limit, Mapping)
+ else None
+ )
+ max_output_tokens = (
+ optional_positive_int(limit.get("output"))
+ if isinstance(limit, Mapping)
+ else None
+ )
effective_package = model_package or provider_package or _DEFAULT_PACKAGE
route = OpenCodeModelRoute(
selector_id=selector_id,
@@ -138,6 +152,8 @@ def parse_open_code_catalog(
),
supports_thinking=supports_thinking,
input_modalities=input_modalities,
+ context_window_tokens=context_window_tokens,
+ max_output_tokens=max_output_tokens,
)
routes[selector_id] = route
@@ -148,6 +164,8 @@ def parse_open_code_catalog(
model_id=route.selector_id,
supports_thinking=route.supports_thinking,
input_modalities=route.input_modalities,
+ context_window_tokens=route.context_window_tokens,
+ max_output_tokens=route.max_output_tokens,
)
for route in routes.values()
)
diff --git a/tests/api/test_model_listing.py b/tests/api/test_model_listing.py
index 09fed0e5..6e7c56d7 100644
--- a/tests/api/test_model_listing.py
+++ b/tests/api/test_model_listing.py
@@ -237,11 +237,14 @@ def test_direct_model_views_serialize_known_capabilities_and_omit_unknowns():
input_modalities=frozenset(
{ModelInputModality.TEXT, ModelInputModality.IMAGE}
),
+ context_window_tokens=131072,
+ max_output_tokens=8192,
),
ProviderModelInfo(
"text-only",
supports_thinking=False,
input_modalities=frozenset({ModelInputModality.TEXT}),
+ context_window_tokens=65536,
),
ProviderModelInfo("unknown"),
},
@@ -262,12 +265,20 @@ def test_direct_model_views_serialize_known_capabilities_and_omit_unknowns():
"text",
"image",
]
+ assert messages["open_router/vision-reasoning"]["contextWindow"] == 131072
+ assert messages["open_router/vision-reasoning"]["maxCompletionTokens"] == 8192
assert messages["open_router/text-only"]["supportsReasoning"] is False
assert messages["open_router/text-only"]["inputModalities"] == ["text"]
+ assert messages["open_router/text-only"]["contextWindow"] == 65536
+ assert "maxCompletionTokens" not in messages["open_router/text-only"]
assert "supportsReasoning" not in messages["open_router/unknown"]
assert "inputModalities" not in messages["open_router/unknown"]
+ assert "contextWindow" not in messages["open_router/unknown"]
+ assert "maxCompletionTokens" not in messages["open_router/unknown"]
assert responses["open_router/text-only"]["supportsReasoningEffort"] is False
assert "reasoningEfforts" not in responses["open_router/text-only"]
+ assert responses["open_router/vision-reasoning"]["contextWindow"] == 131072
+ assert responses["open_router/vision-reasoning"]["maxCompletionTokens"] == 8192
def test_claude_model_view_does_not_expose_capability_fields():
@@ -281,6 +292,8 @@ def test_claude_model_view_does_not_expose_capability_fields():
input_modalities=frozenset(
{ModelInputModality.TEXT, ModelInputModality.IMAGE}
),
+ context_window_tokens=131072,
+ max_output_tokens=8192,
)
},
)
@@ -288,7 +301,11 @@ def test_claude_model_view_does_not_expose_capability_fields():
rows = TestClient(app).get("/v1/models").json()["data"]
assert all(
- "supportsReasoning" not in row and "inputModalities" not in row for row in rows
+ "supportsReasoning" not in row
+ and "inputModalities" not in row
+ and "contextWindow" not in row
+ and "maxCompletionTokens" not in row
+ for row in rows
)
diff --git a/tests/application/test_model_metadata.py b/tests/application/test_model_metadata.py
index 7fce6e84..c8b9bcfa 100644
--- a/tests/application/test_model_metadata.py
+++ b/tests/application/test_model_metadata.py
@@ -9,10 +9,14 @@ def test_provider_model_info_preserves_capabilities_when_identity_is_replaced()
model_id="vendor/model",
supports_thinking=True,
input_modalities=frozenset({ModelInputModality.TEXT, ModelInputModality.IMAGE}),
+ context_window_tokens=131072,
+ max_output_tokens=8192,
)
assert replace(info, model_id="provider/vendor/model") == ProviderModelInfo(
model_id="provider/vendor/model",
supports_thinking=True,
input_modalities=frozenset({ModelInputModality.TEXT, ModelInputModality.IMAGE}),
+ context_window_tokens=131072,
+ max_output_tokens=8192,
)
diff --git a/tests/cli/test_aider_config.py b/tests/cli/test_aider_config.py
index 407efb1d..ad86c571 100644
--- a/tests/cli/test_aider_config.py
+++ b/tests/cli/test_aider_config.py
@@ -19,6 +19,8 @@ def _models() -> tuple[ClientModel, ...]:
input_modalities=frozenset(
{ModelInputModality.TEXT, ModelInputModality.IMAGE}
),
+ context_window_tokens=131072,
+ max_output_tokens=8192,
),
ClientModel(
wire_slug="ollama_cloud/qwen3-coder:480b",
@@ -26,6 +28,7 @@ def _models() -> tuple[ClientModel, ...]:
display_name="Colon model",
supports_reasoning=False,
input_modalities=frozenset({ModelInputModality.TEXT}),
+ max_output_tokens=4096,
),
ClientModel(
wire_slug="future_provider/unknown-model",
@@ -70,11 +73,14 @@ def test_aider_config_projects_messages_route_and_canonical_catalog() -> None:
"litellm_provider": "anthropic",
"mode": "chat",
"supports_vision": True,
+ "max_input_tokens": 131072,
+ "max_output_tokens": 8192,
},
"anthropic/ollama_cloud/qwen3-coder:480b": {
"litellm_provider": "anthropic",
"mode": "chat",
"supports_vision": False,
+ "max_output_tokens": 4096,
},
"anthropic/future_provider/unknown-model": {
"litellm_provider": "anthropic",
diff --git a/tests/cli/test_cline_launcher.py b/tests/cli/test_cline_launcher.py
index 32f68c69..ba3f9a3c 100644
--- a/tests/cli/test_cline_launcher.py
+++ b/tests/cli/test_cline_launcher.py
@@ -59,6 +59,8 @@ def test_cline_config_uses_responses_and_only_known_metadata() -> None:
input_modalities=frozenset(
{ModelInputModality.TEXT, ModelInputModality.IMAGE}
),
+ context_window_tokens=131072,
+ max_output_tokens=8192,
),
ClientModel(
wire_slug="claude-3-freecc-no-thinking/open_router/plain-model",
@@ -66,6 +68,7 @@ def test_cline_config_uses_responses_and_only_known_metadata() -> None:
display_name="No-thinking model",
supports_reasoning=False,
input_modalities=frozenset({ModelInputModality.TEXT}),
+ context_window_tokens=65536,
),
ClientModel(
wire_slug="future_provider/unknown-model",
@@ -123,6 +126,8 @@ def test_cline_config_uses_responses_and_only_known_metadata() -> None:
"supportsVision": True,
"inputModalities": ["text", "image"],
"outputModalities": ["text"],
+ "contextWindow": 131072,
+ "maxTokens": 8192,
"apiFormat": "openai-responses",
},
"claude-3-freecc-no-thinking/open_router/plain-model": {
@@ -132,6 +137,7 @@ def test_cline_config_uses_responses_and_only_known_metadata() -> None:
"supportsVision": False,
"inputModalities": ["text"],
"outputModalities": ["text"],
+ "contextWindow": 65536,
"apiFormat": "openai-responses",
},
"future_provider/unknown-model": {
@@ -145,8 +151,6 @@ def test_cline_config_uses_responses_and_only_known_metadata() -> None:
},
}
serialized = json.dumps(config.providers | config.models)
- assert "contextWindow" not in serialized
- assert "maxTokens" not in serialized
assert "reasoningEffort" not in serialized
assert "proxy-token" not in repr(config)
diff --git a/tests/cli/test_codex_model_catalog.py b/tests/cli/test_codex_model_catalog.py
index 8e8d3ecf..65a8245d 100644
--- a/tests/cli/test_codex_model_catalog.py
+++ b/tests/cli/test_codex_model_catalog.py
@@ -159,12 +159,15 @@ def test_codex_catalog_projects_known_capabilities_and_preserves_unknown_default
"provider_model_ref": "provider/vision-reasoning",
"supportsReasoning": True,
"inputModalities": ["text", "image"],
+ "contextWindow": 131072,
+ "maxCompletionTokens": 8192,
},
{
"id": "claude-3-freecc-no-thinking/provider/text-only",
"provider_model_ref": "provider/text-only",
"supportsReasoning": False,
"inputModalities": ["text"],
+ "maxCompletionTokens": 4096,
},
{
"id": "provider/unknown",
@@ -180,18 +183,24 @@ def test_codex_catalog_projects_known_capabilities_and_preserves_unknown_default
assert vision["supported_reasoning_levels"]
assert vision["supports_reasoning_summaries"] is True
assert vision["default_reasoning_summary"] == "none"
+ assert vision["context_window"] == 131072
+ assert vision["max_context_window"] == 131072
assert text_only["input_modalities"] == ["text"]
assert "default_reasoning_level" not in text_only
assert text_only["supported_reasoning_levels"] == []
assert text_only["supports_reasoning_summaries"] is False
assert "default_reasoning_summary" not in text_only
+ assert text_only["context_window"] == 200000
+ assert text_only["max_context_window"] == 200000
assert unknown["input_modalities"] == ["text"]
assert unknown["default_reasoning_level"] == "medium"
assert unknown["supported_reasoning_levels"]
assert unknown["supports_reasoning_summaries"] is True
assert unknown["default_reasoning_summary"] == "none"
+ assert unknown["context_window"] == 200000
+ assert unknown["max_context_window"] == 200000
def test_launcher_config_composes_with_persistent_codex_config(
diff --git a/tests/cli/test_dsh_config.py b/tests/cli/test_dsh_config.py
index b68d9d5f..62a7d943 100644
--- a/tests/cli/test_dsh_config.py
+++ b/tests/cli/test_dsh_config.py
@@ -21,6 +21,8 @@ def _models() -> tuple[ClientModel, ...]:
input_modalities=frozenset(
{ModelInputModality.TEXT, ModelInputModality.IMAGE}
),
+ context_window_tokens=131072,
+ max_output_tokens=8192,
),
ClientModel(
wire_slug="claude-3-freecc-no-thinking/open_router/plain-model",
@@ -28,6 +30,7 @@ def _models() -> tuple[ClientModel, ...]:
display_name="No-thinking model",
supports_reasoning=False,
input_modalities=frozenset({ModelInputModality.TEXT}),
+ max_output_tokens=4096,
),
ClientModel(
wire_slug="future_provider/unknown-model",
@@ -92,12 +95,15 @@ def test_dsh_config_pins_responses_models_retries_and_private_state(
"max": "max",
},
"input": ["text", "image"],
+ "contextWindow": 131072,
+ "maxTokens": 8192,
},
{
"id": "claude-3-freecc-no-thinking/open_router/plain-model",
"name": "No-thinking model",
"reasoningEfforts": False,
"input": ["text"],
+ "maxTokens": 4096,
},
{
"id": "future_provider/unknown-model",
@@ -139,13 +145,7 @@ def test_dsh_config_pins_responses_models_retries_and_private_state(
serialized = json.dumps(launch.patch)
assert "proxy-token" not in serialized
- for unsupported in (
- "contextWindow",
- "maxTokens",
- "compat",
- "headers",
- "telemetry",
- ):
+ for unsupported in ("compat", "headers", "telemetry"):
assert unsupported not in serialized
diff --git a/tests/cli/test_hermes_config.py b/tests/cli/test_hermes_config.py
index 68a22b18..35adf90d 100644
--- a/tests/cli/test_hermes_config.py
+++ b/tests/cli/test_hermes_config.py
@@ -21,6 +21,8 @@ def _models() -> tuple[ClientModel, ...]:
input_modalities=frozenset(
{ModelInputModality.TEXT, ModelInputModality.IMAGE}
),
+ context_window_tokens=131072,
+ max_output_tokens=8192,
),
ClientModel(
wire_slug="claude-3-freecc-no-thinking/open_router/plain-model",
@@ -28,6 +30,7 @@ def _models() -> tuple[ClientModel, ...]:
display_name="No-thinking model",
supports_reasoning=False,
input_modalities=frozenset({ModelInputModality.TEXT}),
+ max_output_tokens=4096,
),
ClientModel(
wire_slug="future_provider/unknown-model",
@@ -81,10 +84,13 @@ def test_hermes_config_pins_responses_catalog_and_fallbacks() -> None:
"nvidia_nim/vendor/model": {
"supports_reasoning": True,
"supports_vision": True,
+ "context_window": 131072,
+ "max_output_tokens": 8192,
},
"claude-3-freecc-no-thinking/open_router/plain-model": {
"supports_reasoning": False,
"supports_vision": False,
+ "max_output_tokens": 4096,
},
}
}
diff --git a/tests/cli/test_model_catalog.py b/tests/cli/test_model_catalog.py
index a05e61ff..2bf23e58 100644
--- a/tests/cli/test_model_catalog.py
+++ b/tests/cli/test_model_catalog.py
@@ -101,12 +101,16 @@ def test_client_models_parse_capabilities_without_deriving_reasoning_from_slug()
"provider_model_ref": "provider/reasoning",
"supportsReasoning": False,
"inputModalities": ["text"],
+ "contextWindow": 131072,
+ "maxCompletionTokens": 8192,
},
{
"id": "claude-3-freecc-no-thinking/provider/unknown",
"provider_model_ref": "provider/unknown",
"supportsReasoning": "not-a-bool",
"inputModalities": ["text", "image"],
+ "contextWindow": 0,
+ "maxCompletionTokens": "8192",
},
{
"id": "provider/malformed-media",
@@ -118,13 +122,23 @@ def test_client_models_parse_capabilities_without_deriving_reasoning_from_slug()
}
)
- assert [(model.supports_reasoning, model.input_modalities) for model in models] == [
- (False, frozenset({ModelInputModality.TEXT})),
+ assert [
+ (
+ model.supports_reasoning,
+ model.input_modalities,
+ model.context_window_tokens,
+ model.max_output_tokens,
+ )
+ for model in models
+ ] == [
+ (False, frozenset({ModelInputModality.TEXT}), 131072, 8192),
(
None,
frozenset({ModelInputModality.TEXT, ModelInputModality.IMAGE}),
+ None,
+ None,
),
- (True, None),
+ (True, None, None, None),
]
diff --git a/tests/cli/test_opencode_launcher.py b/tests/cli/test_opencode_launcher.py
index dac1cd52..c0190f8f 100644
--- a/tests/cli/test_opencode_launcher.py
+++ b/tests/cli/test_opencode_launcher.py
@@ -54,6 +54,8 @@ def test_opencode_config_uses_responses_sdk_and_only_known_metadata() -> None:
input_modalities=frozenset(
{ModelInputModality.TEXT, ModelInputModality.IMAGE}
),
+ context_window_tokens=131072,
+ max_output_tokens=8192,
),
ClientModel(
wire_slug="claude-3-freecc-no-thinking/open_router/plain-model",
@@ -61,6 +63,7 @@ def test_opencode_config_uses_responses_sdk_and_only_known_metadata() -> None:
display_name="No-thinking model",
supports_reasoning=False,
input_modalities=frozenset({ModelInputModality.TEXT}),
+ context_window_tokens=65536,
),
ClientModel(
wire_slug="future_provider/unknown-model",
@@ -68,6 +71,13 @@ def test_opencode_config_uses_responses_sdk_and_only_known_metadata() -> None:
display_name="Unknown model",
supports_reasoning=None,
),
+ ClientModel(
+ wire_slug="future_provider/output-only",
+ provider_model_ref="future_provider/output-only",
+ display_name="Output-only model",
+ supports_reasoning=None,
+ max_output_tokens=4096,
+ ),
),
proxy_root_url="http://127.0.0.1:9191",
)
@@ -86,16 +96,23 @@ def test_opencode_config_uses_responses_sdk_and_only_known_metadata() -> None:
"name": "Nested model",
"reasoning": True,
"modalities": {"input": ["text", "image"]},
+ "limit": {"context": 131072, "output": 8192},
},
"claude-3-freecc-no-thinking/open_router/plain-model": {
"name": "No-thinking model",
"reasoning": False,
"modalities": {"input": ["text"]},
+ "limit": {"context": 65536, "output": 0},
},
"future_provider/unknown-model": {
"name": "Unknown model",
"reasoning": True,
},
+ "future_provider/output-only": {
+ "name": "Output-only model",
+ "reasoning": True,
+ "limit": {"context": 0, "output": 4096},
+ },
}
assert config.overlay == {
"provider": {
@@ -115,7 +132,6 @@ def test_opencode_config_uses_responses_sdk_and_only_known_metadata() -> None:
}
serialized = json.dumps(config.file | config.overlay)
assert "proxy-token" not in serialized
- assert "context" not in serialized
assert "attachment" not in serialized
diff --git a/tests/cli/test_pi_extension.py b/tests/cli/test_pi_extension.py
index 182fb031..47b7ae74 100644
--- a/tests/cli/test_pi_extension.py
+++ b/tests/cli/test_pi_extension.py
@@ -24,12 +24,15 @@ def test_pi_extension_projects_known_capabilities_and_preserves_unknown_defaults
"provider_model_ref": "provider/vision-reasoning",
"supportsReasoning": True,
"inputModalities": ["text", "image"],
+ "contextWindow": 131072,
+ "maxCompletionTokens": 8192,
},
{
"id": "claude-3-freecc-no-thinking/provider/text-only",
"provider_model_ref": "provider/text-only",
"supportsReasoning": False,
"inputModalities": ["text"],
+ "contextWindow": 65536,
},
{
"id": "provider/unknown",
@@ -68,3 +71,8 @@ console.log(JSON.stringify(projectFccModels(payload)));
(False, ["text"]),
(True, ["text"]),
]
+ assert [(model["contextWindow"], model["maxTokens"]) for model in projected] == [
+ (131072, 8192),
+ (65536, 16384),
+ (128000, 16384),
+ ]
diff --git a/tests/providers/test_groq.py b/tests/providers/test_groq.py
index 985dc626..87075b94 100644
--- a/tests/providers/test_groq.py
+++ b/tests/providers/test_groq.py
@@ -4,6 +4,7 @@ from unittest.mock import AsyncMock, MagicMock, patch
import pytest
+from free_claude_code.application.model_metadata import ProviderModelInfo
from free_claude_code.config.provider_catalog import GROQ_DEFAULT_BASE
from free_claude_code.providers.groq import GroqProvider
from tests.providers.request_factory import make_messages_request
@@ -47,6 +48,36 @@ def test_default_base_url_constant():
assert GROQ_DEFAULT_BASE == "https://api.groq.com/openai/v1"
+async def test_model_catalog_extracts_documented_token_limits(groq_provider) -> None:
+ with patch.object(
+ groq_provider,
+ "_list_models_payload",
+ new=AsyncMock(
+ return_value={
+ "data": [
+ {
+ "id": "model",
+ "context_window": 131072,
+ "max_completion_tokens": 8192,
+ }
+ ]
+ }
+ ),
+ ):
+ infos = await groq_provider.list_model_infos()
+
+ assert infos == frozenset(
+ {
+ ProviderModelInfo(
+ "model",
+ context_window_tokens=131072,
+ max_output_tokens=8192,
+ )
+ }
+ )
+
+
def test_build_request_body_basic(groq_provider):
"""Basic request body conversion attaches system message from Claude request."""
req = make_request()
diff --git a/tests/providers/test_huggingface.py b/tests/providers/test_huggingface.py
index 9a39e100..fb4417dd 100644
--- a/tests/providers/test_huggingface.py
+++ b/tests/providers/test_huggingface.py
@@ -78,6 +78,11 @@ async def test_model_catalog_extracts_exact_input_modalities(
{
"id": "vision-model",
"architecture": {"input_modalities": ["text", "image", "audio"]},
+ "providers": [
+ {"status": "live", "context_length": 131072},
+ {"status": "live", "context_length": 131072},
+ {"status": "staging", "context_length": 999999},
+ ],
},
{
"id": "unknown-model",
@@ -94,12 +99,40 @@ async def test_model_catalog_extracts_exact_input_modalities(
input_modalities=frozenset(
{ModelInputModality.TEXT, ModelInputModality.IMAGE}
),
+ context_window_tokens=131072,
),
ProviderModelInfo("unknown-model"),
}
)
+ "providers",
+ [
+ [],
+ [{"status": "staging", "context_length": 131072}],
+ [{"status": "live"}],
+ [{"status": "live", "context_length": "131072"}],
+ [
+ {"status": "live", "context_length": 131072},
+ {"status": "live", "context_length": 65536},
+ ],
+ ],
+)
+async def test_model_catalog_requires_consensus_across_live_huggingface_routes(
+ huggingface_provider,
+ providers: list[dict[str, object]],
+) -> None:
+ huggingface_provider._client.models.list = AsyncMock(
+ return_value=SimpleNamespace(data=[{"id": "model", "providers": providers}])
+ )
+
+ assert await huggingface_provider.list_model_infos() == frozenset(
+ {ProviderModelInfo("model")}
+ )
+
+
def test_build_request_body_keeps_max_tokens(huggingface_provider):
with patch(
"free_claude_code.providers.openai_chat.request_policy.build_base_request_body"
diff --git a/tests/providers/test_kilo.py b/tests/providers/test_kilo.py
index 11749c6e..74e88ffe 100644
--- a/tests/providers/test_kilo.py
+++ b/tests/providers/test_kilo.py
@@ -380,6 +380,8 @@ async def test_model_list_filters_to_chat_tool_models_with_capabilities(kilo_pro
"output_modalities": ["text"],
},
"opencode": {"ai_sdk_provider": "anthropic"},
+ "context_length": 200000,
+ "top_provider": {"max_completion_tokens": 16000},
},
{
"id": "plain-tool",
@@ -426,6 +428,8 @@ async def test_model_list_filters_to_chat_tool_models_with_capabilities(kilo_pro
input_modalities=frozenset(
{ModelInputModality.TEXT, ModelInputModality.IMAGE}
),
+ context_window_tokens=200000,
+ max_output_tokens=16000,
),
ProviderModelInfo("plain-tool", supports_thinking=False),
ProviderModelInfo(
diff --git a/tests/providers/test_llm7.py b/tests/providers/test_llm7.py
index 09524b0b..48da2fd7 100644
--- a/tests/providers/test_llm7.py
+++ b/tests/providers/test_llm7.py
@@ -71,6 +71,7 @@ def _catalog_model(
tools_calling: object = True,
reasoning: bool | None = None,
input_modalities: object = ("text",),
+ context_window_tokens: int | None = None,
) -> dict[str, object]:
model: dict[str, object] = {
"id": model_id,
@@ -81,6 +82,11 @@ def _catalog_model(
}
if reasoning is not None:
model["reasoning"] = reasoning
+ if context_window_tokens is not None:
+ model["context_window"] = {
+ "tokens": context_window_tokens,
+ "characters": context_window_tokens * 4,
+ }
return model
@@ -185,7 +191,11 @@ async def test_filters_catalog_and_adds_live_authoritative_selectors(
),
_catalog_model("plain-model", reasoning=False),
_catalog_model("unknown-reasoning-model"),
- _catalog_model("default", reasoning=True),
+ _catalog_model(
+ "default",
+ reasoning=True,
+ context_window_tokens=114688,
+ ),
_catalog_model("image-generator", model_type="image"),
_catalog_model("video-generator", model_type="video"),
_catalog_model("nonstreaming-chat", stream=False),
diff --git a/tests/providers/test_mistral.py b/tests/providers/test_mistral.py
index 346eaa59..f8e7cc32 100644
--- a/tests/providers/test_mistral.py
+++ b/tests/providers/test_mistral.py
@@ -62,6 +62,7 @@ async def test_model_catalog_extracts_exact_input_modalities(mistral_provider):
"data": [
{
"id": "vision-model",
+ "max_context_length": 131072,
"capabilities": {
"completion_chat": True,
"vision": True,
@@ -85,6 +86,7 @@ async def test_model_catalog_extracts_exact_input_modalities(mistral_provider):
input_modalities=frozenset(
{ModelInputModality.TEXT, ModelInputModality.IMAGE}
),
+ context_window_tokens=131072,
),
ProviderModelInfo(
"text-model",
diff --git a/tests/providers/test_model_token_limit_profiles.py b/tests/providers/test_model_token_limit_profiles.py
new file mode 100644
index 00000000..b2b5ff33
--- /dev/null
+++ b/tests/providers/test_model_token_limit_profiles.py
@@ -0,0 +1,177 @@
+"""Provider-profile contracts for advertised model token limits."""
+
+from unittest.mock import AsyncMock, patch
+
+import pytest
+
+from free_claude_code.application.model_metadata import ProviderModelInfo
+from tests.providers.support import (
+ immediate_admission,
+ make_provider_config,
+ profiled_provider,
+)
+
+
+ ("provider_id", "payload", "context_window_tokens", "max_output_tokens"),
+ [
+ (
+ "vercel",
+ {"data": [{"id": "model", "context_window": 131072, "max_tokens": 8192}]},
+ 131072,
+ 8192,
+ ),
+ (
+ "chutes",
+ {
+ "data": [
+ {
+ "id": "model",
+ "input_modalities": ["text"],
+ "output_modalities": ["text"],
+ "supported_features": ["tools"],
+ "context_length": 32768,
+ "max_output_length": 4096,
+ }
+ ]
+ },
+ 32768,
+ 4096,
+ ),
+ (
+ "featherless",
+ {
+ "data": [
+ {
+ "id": "model",
+ "features": {"tool_use": True, "image_input": False},
+ "is_gated": False,
+ "available_on_current_plan": True,
+ "context_length": 65536,
+ "max_completion_tokens": 8192,
+ }
+ ]
+ },
+ 65536,
+ 8192,
+ ),
+ (
+ "zenmux",
+ {
+ "data": [
+ {
+ "id": "model",
+ "input_modalities": ["text"],
+ "output_modalities": ["text"],
+ "context_length": 200000,
+ }
+ ]
+ },
+ 200000,
+ None,
+ ),
+ (
+ "deepinfra",
+ [
+ {
+ "model_name": "model",
+ "reported_type": "text-generation",
+ "deprecated": None,
+ "max_tokens": 131072,
+ }
+ ],
+ 131072,
+ None,
+ ),
+ (
+ "nebius",
+ {
+ "data": [
+ {
+ "id": "model",
+ "architecture": {"modality": "text->text"},
+ "context_length": 128000,
+ }
+ ]
+ },
+ 128000,
+ None,
+ ),
+ (
+ "mistral_codestral",
+ {
+ "data": [
+ {
+ "id": "model",
+ "capabilities": {
+ "completion_chat": True,
+ "vision": False,
+ },
+ "max_context_length": 256000,
+ }
+ ]
+ },
+ 256000,
+ None,
+ ),
+ (
+ "together",
+ [{"id": "model", "type": "chat", "context_length": 131072}],
+ 131072,
+ None,
+ ),
+ (
+ "llm7",
+ {
+ "data": [
+ {
+ "id": "model",
+ "model_type": "chat",
+ "stream": True,
+ "tools_calling": True,
+ "context_window": {"tokens": 114688, "characters": 458752},
+ }
+ ]
+ },
+ 114688,
+ None,
+ ),
+ ],
+)
+async def test_profile_extracts_only_its_documented_token_limit_fields(
+ provider_id: str,
+ payload: object,
+ context_window_tokens: int | None,
+ max_output_tokens: int | None,
+) -> None:
+ provider = profiled_provider(
+ provider_id,
+ make_provider_config(api_key="test", base_url="https://provider.test/v1"),
+ admission=immediate_admission(provider_name=provider_id),
+ )
+ try:
+ with patch.object(
+ provider,
+ "_list_models_payload",
+ new=AsyncMock(return_value=payload),
+ ):
+ infos = await provider.list_model_infos()
+ finally:
+ await provider.cleanup()
+
+ info = next(info for info in infos if info.model_id == "model")
+ assert info == ProviderModelInfo(
+ "model",
+ supports_thinking=info.supports_thinking,
+ input_modalities=info.input_modalities,
+ context_window_tokens=context_window_tokens,
+ max_output_tokens=max_output_tokens,
+ )
+ if provider_id == "llm7":
+ selectors = {info.model_id: info for info in infos if info.model_id != "model"}
+ assert set(selectors) == {"default", "fast", "pro"}
+ assert all(
+ info.context_window_tokens is None and info.max_output_tokens is None
+ for info in selectors.values()
+ )
diff --git a/tests/providers/test_open_router.py b/tests/providers/test_open_router.py
index e31ece72..2f728d92 100644
--- a/tests/providers/test_open_router.py
+++ b/tests/providers/test_open_router.py
@@ -344,6 +344,11 @@ async def test_model_infos_filter_tool_models_and_thinking_metadata(
id="tool-model",
supported_parameters=["tools", "reasoning"],
architecture=SimpleNamespace(input_modalities=["text", "image"]),
+ context_length=262144,
+ top_provider=SimpleNamespace(
+ context_length=999999,
+ max_completion_tokens=32768,
+ ),
),
SimpleNamespace(
id="tool-model-with-malformed-capabilities",
@@ -365,6 +370,8 @@ async def test_model_infos_filter_tool_models_and_thinking_metadata(
input_modalities=frozenset(
{ModelInputModality.TEXT, ModelInputModality.IMAGE}
),
+ context_window_tokens=262144,
+ max_output_tokens=32768,
),
ProviderModelInfo(
"tool-model-with-malformed-capabilities",
diff --git a/tests/providers/test_openai_codex_provider.py b/tests/providers/test_openai_codex_provider.py
index 5606f57d..8b57477f 100644
--- a/tests/providers/test_openai_codex_provider.py
+++ b/tests/providers/test_openai_codex_provider.py
@@ -197,6 +197,7 @@ async def test_provider_uses_subscription_headers_and_visible_model_catalog() ->
"supported_in_api": False,
"supported_reasoning_levels": [{"effort": "high"}],
"input_modalities": ["text", "image"],
+ "context_window": 272000,
},
{
"slug": "gpt-no-reasoning",
@@ -248,6 +249,7 @@ async def test_provider_uses_subscription_headers_and_visible_model_catalog() ->
input_modalities=frozenset(
{ModelInputModality.TEXT, ModelInputModality.IMAGE}
),
+ context_window_tokens=272000,
),
ProviderModelInfo("gpt-no-reasoning", supports_thinking=False),
ProviderModelInfo("gpt-malformed-reasoning"),
diff --git a/tests/providers/test_openai_model_listing_filters.py b/tests/providers/test_openai_model_listing_filters.py
index 98c292a7..7aa9c5e5 100644
--- a/tests/providers/test_openai_model_listing_filters.py
+++ b/tests/providers/test_openai_model_listing_filters.py
@@ -99,6 +99,7 @@ def test_optional_model_capabilities_are_normalized_and_copied_to_aliases() -> N
"aliases": ["latest"],
"architecture": {"input_modalities": ["text", "image", "audio"]},
"supported_parameters": ["tools", "reasoning"],
+ "limits": {"context": 131072, "output": 16384},
},
{
"id": "text-only",
@@ -112,12 +113,16 @@ def test_optional_model_capabilities_are_normalized_and_copied_to_aliases() -> N
aliases_field="aliases",
input_modalities_path=("architecture", "input_modalities"),
thinking_sequence_path=("supported_parameters",),
+ context_window_tokens_path=("limits", "context"),
+ max_output_tokens_path=("limits", "output"),
)
vision = ProviderModelInfo(
"vision-reasoning",
supports_thinking=True,
input_modalities=frozenset({ModelInputModality.TEXT, ModelInputModality.IMAGE}),
+ context_window_tokens=131072,
+ max_output_tokens=16384,
)
assert model_infos == frozenset(
{
@@ -126,6 +131,8 @@ def test_optional_model_capabilities_are_normalized_and_copied_to_aliases() -> N
"latest",
supports_thinking=True,
input_modalities=vision.input_modalities,
+ context_window_tokens=131072,
+ max_output_tokens=16384,
),
ProviderModelInfo(
"text-only",
@@ -136,6 +143,50 @@ def test_optional_model_capabilities_are_normalized_and_copied_to_aliases() -> N
)
+ "invalid_value",
+ [True, False, 0, -1, 1.5, "4096", None, [], {}],
+)
+def test_optional_token_limits_require_exact_positive_integers(
+ invalid_value: object,
+) -> None:
+ [info] = extract_openai_model_infos(
+ {
+ "data": [
+ {
+ "id": "model",
+ "limits": {"context": invalid_value, "output": invalid_value},
+ }
+ ]
+ },
+ provider_name="TEST",
+ context_window_tokens_path=("limits", "context"),
+ max_output_tokens_path=("limits", "output"),
+ )
+
+ assert info.context_window_tokens is None
+ assert info.max_output_tokens is None
+
+
+def test_optional_token_limits_degrade_independently() -> None:
+ [info] = extract_openai_model_infos(
+ {
+ "data": [
+ {
+ "id": "model",
+ "limits": {"context": "bad", "output": 8192},
+ }
+ ]
+ },
+ provider_name="TEST",
+ context_window_tokens_path=("limits", "context"),
+ max_output_tokens_path=("limits", "output"),
+ )
+
+ assert info.context_window_tokens is None
+ assert info.max_output_tokens == 8192
+
+
@pytest.mark.parametrize(
("field", "value", "expected"),
[
@@ -202,6 +253,8 @@ def test_tool_capable_parser_retains_exact_input_modalities() -> None:
"id": "vision",
"supported_parameters": ["tools", "reasoning"],
"architecture": {"input_modalities": ["text", "image", "audio"]},
+ "context_length": 262144,
+ "top_provider": {"max_completion_tokens": 32768},
},
{
"id": "unknown-media",
@@ -221,6 +274,8 @@ def test_tool_capable_parser_retains_exact_input_modalities() -> None:
input_modalities=frozenset(
{ModelInputModality.TEXT, ModelInputModality.IMAGE}
),
+ context_window_tokens=262144,
+ max_output_tokens=32768,
),
ProviderModelInfo("unknown-media", supports_thinking=False),
}
diff --git a/tests/providers/test_opencode.py b/tests/providers/test_opencode.py
index fd4f78a0..41581f49 100644
--- a/tests/providers/test_opencode.py
+++ b/tests/providers/test_opencode.py
@@ -329,6 +329,7 @@ def test_catalog_resolves_package_precedence_status_alias_and_reasoning() -> Non
"id": "provider-default",
"reasoning": False,
"modalities": {"input": ["text"]},
+ "limit": {"context": 131072, "output": 8192},
},
"responses-alias": {
"id": "actual-responses-id",
@@ -382,6 +383,8 @@ def test_catalog_resolves_package_precedence_status_alias_and_reasoning() -> Non
"provider-default",
supports_thinking=False,
input_modalities=frozenset({ModelInputModality.TEXT}),
+ context_window_tokens=131072,
+ max_output_tokens=8192,
),
ProviderModelInfo(
"responses-alias",
@@ -412,6 +415,28 @@ def test_catalog_defaults_missing_package_to_chat_completions() -> None:
assert route.transport is OpenCodeUpstreamTransport.CHAT_COMPLETIONS
+def test_catalog_token_limits_degrade_independently_without_rejecting_model() -> None:
+ snapshot = parse_open_code_catalog(
+ _catalog_payload(
+ {
+ "partial": {
+ "id": "upstream",
+ "limit": {"context": "bad", "output": 8192},
+ },
+ "malformed": {"id": "other", "limit": "bad"},
+ }
+ ),
+ provider_key="opencode",
+ provider_name="OPENCODE_ZEN",
+ )
+
+ infos = {info.model_id: info for info in snapshot.model_infos}
+ assert infos["partial"].context_window_tokens is None
+ assert infos["partial"].max_output_tokens == 8192
+ assert infos["malformed"].context_window_tokens is None
+ assert infos["malformed"].max_output_tokens is None
+
+
@pytest.mark.parametrize(
"payload,match",
[
diff --git a/tests/runtime/test_model_cache.py b/tests/runtime/test_model_cache.py
index 477f1fda..4423d662 100644
--- a/tests/runtime/test_model_cache.py
+++ b/tests/runtime/test_model_cache.py
@@ -9,6 +9,8 @@ def test_model_cache_returns_and_prefixes_complete_model_metadata() -> None:
model_id="vendor/model",
supports_thinking=False,
input_modalities=frozenset({ModelInputModality.TEXT, ModelInputModality.IMAGE}),
+ context_window_tokens=131072,
+ max_output_tokens=8192,
)
cache.cache_model_infos("open_router", (info,))
@@ -22,5 +24,7 @@ def test_model_cache_returns_and_prefixes_complete_model_metadata() -> None:
input_modalities=frozenset(
{ModelInputModality.TEXT, ModelInputModality.IMAGE}
),
+ context_window_tokens=131072,
+ max_output_tokens=8192,
),
)
diff --git a/uv.lock b/uv.lock
index 2033b1e7..61081281 100644
--- a/uv.lock
+++ b/uv.lock
@@ -607,7 +607,7 @@ wheels = [
[[package]]
name = "free-claude-code"
-version = "5.16.0"
+version = "5.17.0"
source = { editable = "." }
dependencies = [
{ name = "aiohttp" },