Skip to content

Drop in a screenshot and convert it to clean code (HTML/Tailwind/React/Vue)

Python

Replace GPT 5.2 Codex with GPT 5.5 models (#605)

Abi Raja260cc55a82fe
diff --git a/backend/agent/providers/pricing.py b/backend/agent/providers/pricing.py
index e33ba226..d3b9d363 100644
--- a/backend/agent/providers/pricing.py
+++ b/backend/agent/providers/pricing.py
@@ -15,9 +15,6 @@ class ModelPricing:
# Pricing keyed by the API model name string sent to the provider.
MODEL_PRICING: Dict[str, ModelPricing] = {
# --- OpenAI ---
- "gpt-5.2-codex": ModelPricing(
- input=1.75, output=14.00, cache_read=0.4375
- ),
"gpt-5.4-mini": ModelPricing(
input=0.40, output=3.20, cache_read=0.10
),
diff --git a/backend/evals/runner.py b/backend/evals/runner.py
index 40acd8a3..bfb5c0d5 100644
--- a/backend/evals/runner.py
+++ b/backend/evals/runner.py
@@ -166,7 +166,7 @@ async def run_image_evals(
if data_url is None:
data_url = await image_to_data_url(filepath)
current_model_for_task = (
- selected_model if n_idx == 0 else Llm.GPT_5_2_CODEX_LOW
+ selected_model if n_idx == 0 else Llm.GPT_5_5_LOW
)
coro = generate_code_and_time(
image_url=data_url,
diff --git a/backend/llm.py b/backend/llm.py
index 318e50b5..685edb52 100644
--- a/backend/llm.py
+++ b/backend/llm.py
@@ -5,10 +5,6 @@ from typing import TypedDict
# Actual model versions that are passed to the LLMs and stored in our logs
class Llm(Enum):
# GPT
- GPT_5_2_CODEX_LOW = "gpt-5.2-codex (low thinking)"
- GPT_5_2_CODEX_MEDIUM = "gpt-5.2-codex (medium thinking)"
- GPT_5_2_CODEX_HIGH = "gpt-5.2-codex (high thinking)"
- GPT_5_2_CODEX_XHIGH = "gpt-5.2-codex (xhigh thinking)"
GPT_5_4_MINI_LOW = "gpt-5.4-mini (low thinking)"
GPT_5_4_2026_03_05_NONE = "gpt-5.4-2026-03-05 (no thinking)"
GPT_5_4_2026_03_05_LOW = "gpt-5.4-2026-03-05 (low thinking)"
@@ -50,10 +46,6 @@ class Completion(TypedDict):
# models elsewhere in the codebase.
MODEL_PROVIDER: dict[Llm, str] = {
# OpenAI models
- Llm.GPT_5_2_CODEX_LOW: "openai",
- Llm.GPT_5_2_CODEX_MEDIUM: "openai",
- Llm.GPT_5_2_CODEX_HIGH: "openai",
- Llm.GPT_5_2_CODEX_XHIGH: "openai",
Llm.GPT_5_4_MINI_LOW: "openai",
Llm.GPT_5_4_2026_03_05_NONE: "openai",
Llm.GPT_5_4_2026_03_05_LOW: "openai",
@@ -91,10 +83,6 @@ ANTHROPIC_MODELS = {m for m, p in MODEL_PROVIDER.items() if p == "anthropic"}
GEMINI_MODELS = {m for m, p in MODEL_PROVIDER.items() if p == "gemini"}
OPENAI_MODEL_CONFIG: dict[Llm, dict[str, str]] = {
- Llm.GPT_5_2_CODEX_LOW: {"api_name": "gpt-5.2-codex", "reasoning_effort": "low"},
- Llm.GPT_5_2_CODEX_MEDIUM: {"api_name": "gpt-5.2-codex", "reasoning_effort": "medium"},
- Llm.GPT_5_2_CODEX_HIGH: {"api_name": "gpt-5.2-codex", "reasoning_effort": "high"},
- Llm.GPT_5_2_CODEX_XHIGH: {"api_name": "gpt-5.2-codex", "reasoning_effort": "xhigh"},
Llm.GPT_5_4_MINI_LOW: {"api_name": "gpt-5.4-mini", "reasoning_effort": "low"},
Llm.GPT_5_4_2026_03_05_NONE: {
"api_name": "gpt-5.4-2026-03-05",
diff --git a/backend/routes/model_choice_sets.py b/backend/routes/model_choice_sets.py
index 186b1a34..38a24c93 100644
--- a/backend/routes/model_choice_sets.py
+++ b/backend/routes/model_choice_sets.py
@@ -12,7 +12,7 @@ VIDEO_VARIANT_MODELS = (
ALL_KEYS_MODELS_DEFAULT = (
Llm.GEMINI_3_FLASH_PREVIEW_MINIMAL,
- Llm.GPT_5_2_CODEX_HIGH,
+ Llm.GPT_5_5_XHIGH,
Llm.GEMINI_3_FLASH_PREVIEW_HIGH,
Llm.GEMINI_3_1_PRO_PREVIEW_HIGH,
)
@@ -21,7 +21,7 @@ ALL_KEYS_MODELS_DEFAULT = (
ALL_KEYS_MODELS_TEXT_CREATE = (
Llm.GEMINI_3_FLASH_PREVIEW_MINIMAL,
- Llm.GPT_5_2_CODEX_HIGH,
+ Llm.GPT_5_5_XHIGH,
Llm.CLAUDE_OPUS_4_6,
Llm.GEMINI_3_1_PRO_PREVIEW_LOW,
)
@@ -44,13 +44,13 @@ GEMINI_ANTHROPIC_MODELS = (
GEMINI_OPENAI_MODELS = (
Llm.GEMINI_3_FLASH_PREVIEW_MINIMAL,
Llm.GEMINI_3_1_PRO_PREVIEW_LOW,
- Llm.GPT_5_2_CODEX_HIGH,
- Llm.GPT_5_2_CODEX_MEDIUM,
+ Llm.GPT_5_5_XHIGH,
+ Llm.GPT_5_5_HIGH,
)
OPENAI_ANTHROPIC_MODELS = (
Llm.CLAUDE_OPUS_4_6,
- Llm.GPT_5_2_CODEX_HIGH,
- Llm.GPT_5_2_CODEX_MEDIUM,
+ Llm.GPT_5_5_XHIGH,
+ Llm.GPT_5_5_HIGH,
)
GEMINI_ONLY_MODELS = (
Llm.GEMINI_3_FLASH_PREVIEW_MINIMAL,
@@ -63,6 +63,6 @@ ANTHROPIC_ONLY_MODELS = (
Llm.CLAUDE_SONNET_4_6,
)
OPENAI_ONLY_MODELS = (
- Llm.GPT_5_2_CODEX_HIGH,
- Llm.GPT_5_2_CODEX_MEDIUM,
+ Llm.GPT_5_5_XHIGH,
+ Llm.GPT_5_5_HIGH,
)
diff --git a/backend/tests/test_model_selection.py b/backend/tests/test_model_selection.py
index 9fe6edc8..2e0fe7b6 100644
--- a/backend/tests/test_model_selection.py
+++ b/backend/tests/test_model_selection.py
@@ -25,7 +25,7 @@ class TestModelSelectionAllKeys:
expected = [
Llm.GEMINI_3_FLASH_PREVIEW_MINIMAL,
- Llm.GPT_5_2_CODEX_HIGH,
+ Llm.GPT_5_5_XHIGH,
Llm.CLAUDE_OPUS_4_6,
Llm.GEMINI_3_1_PRO_PREVIEW_LOW,
]
@@ -44,7 +44,7 @@ class TestModelSelectionAllKeys:
expected = [
Llm.GEMINI_3_FLASH_PREVIEW_MINIMAL,
- Llm.GPT_5_2_CODEX_HIGH,
+ Llm.GPT_5_5_XHIGH,
Llm.GEMINI_3_FLASH_PREVIEW_HIGH,
Llm.GEMINI_3_1_PRO_PREVIEW_HIGH,
]
@@ -129,7 +129,7 @@ class TestModelSelectionOpenAIAnthropic:
@pytest.mark.asyncio
async def test_openai_anthropic(self):
- """OpenAI + Anthropic: Claude Opus 4.6, GPT 5.2 Codex (high/medium), cycling"""
+ """OpenAI + Anthropic: Claude Opus 4.6, GPT 5.5 (xhigh/high), cycling"""
models = await self.model_selector.select_models(
generation_type="create",
input_mode="text",
@@ -140,8 +140,8 @@ class TestModelSelectionOpenAIAnthropic:
expected = [
Llm.CLAUDE_OPUS_4_6,
- Llm.GPT_5_2_CODEX_HIGH,
- Llm.GPT_5_2_CODEX_MEDIUM,
+ Llm.GPT_5_5_XHIGH,
+ Llm.GPT_5_5_HIGH,
Llm.CLAUDE_OPUS_4_6,
]
assert models == expected
@@ -185,7 +185,7 @@ class TestModelSelectionOpenAIOnly:
@pytest.mark.asyncio
async def test_openai_only(self):
- """OpenAI only: GPT 5.2 Codex (high/medium) only"""
+ """OpenAI only: GPT 5.5 (xhigh/high) only"""
models = await self.model_selector.select_models(
generation_type="create",
input_mode="text",
@@ -195,10 +195,10 @@ class TestModelSelectionOpenAIOnly:
)
expected = [
- Llm.GPT_5_2_CODEX_HIGH,
- Llm.GPT_5_2_CODEX_MEDIUM,
- Llm.GPT_5_2_CODEX_HIGH,
- Llm.GPT_5_2_CODEX_MEDIUM,
+ Llm.GPT_5_5_XHIGH,
+ Llm.GPT_5_5_HIGH,
+ Llm.GPT_5_5_XHIGH,
+ Llm.GPT_5_5_HIGH,
]
assert models == expected
diff --git a/backend/tests/test_openai_provider_session.py b/backend/tests/test_openai_provider_session.py
index 552da4e3..523c1bc2 100644
--- a/backend/tests/test_openai_provider_session.py
+++ b/backend/tests/test_openai_provider_session.py
@@ -61,7 +61,7 @@ async def test_openai_provider_session_omits_prompt_cache_key_across_turns() ->
client = _FakeOpenAIClient()
session = OpenAIProviderSession(
client=client, # type: ignore[arg-type]
- model=Llm.GPT_5_2_CODEX_HIGH,
+ model=Llm.GPT_5_5_XHIGH,
prompt_messages=[{"role": "user", "content": "Build a landing page."}],
tools=_test_tools(),
)
@@ -125,19 +125,19 @@ async def test_openai_provider_session_omits_prompt_cache_key_for_all_prompts()
first_session = OpenAIProviderSession(
client=first_client, # type: ignore[arg-type]
- model=Llm.GPT_5_2_CODEX_HIGH,
+ model=Llm.GPT_5_5_XHIGH,
prompt_messages=[{"role": "user", "content": "Build a landing page."}],
tools=_test_tools(),
)
second_session = OpenAIProviderSession(
client=second_client, # type: ignore[arg-type]
- model=Llm.GPT_5_2_CODEX_HIGH,
+ model=Llm.GPT_5_5_XHIGH,
prompt_messages=[{"role": "user", "content": "Build a landing page."}],
tools=_test_tools(),
)
different_prompt_session = OpenAIProviderSession(
client=different_prompt_client, # type: ignore[arg-type]
- model=Llm.GPT_5_2_CODEX_HIGH,
+ model=Llm.GPT_5_5_XHIGH,
prompt_messages=[{"role": "user", "content": "Build a dashboard."}],
tools=_test_tools(),
)
diff --git a/backend/tests/test_token_usage.py b/backend/tests/test_token_usage.py
index c9106f78..4a5d46a1 100644
--- a/backend/tests/test_token_usage.py
+++ b/backend/tests/test_token_usage.py
@@ -281,7 +281,6 @@ class TestAnthropicExtract:
class TestModelPricing:
def test_known_models_have_pricing(self) -> None:
for name in [
- "gpt-5.2-codex",
"claude-opus-4-6",
"claude-sonnet-4-6",
"gemini-3-flash-preview",
@@ -302,5 +301,5 @@ class TestModelPricing:
assert MODEL_PRICING[name].cache_write > 0
def test_openai_gemini_no_cache_write(self) -> None:
- for name in ["gpt-5.2-codex", "gemini-3-flash-preview"]:
+ for name in ["gpt-5.5", "gemini-3-flash-preview"]:
assert MODEL_PRICING[name].cache_write == 0.0
diff --git a/frontend/src/lib/models.ts b/frontend/src/lib/models.ts
index 5fa690eb..840a4f02 100644
--- a/frontend/src/lib/models.ts
+++ b/frontend/src/lib/models.ts
@@ -13,10 +13,6 @@ export enum CodeGenerationModel {
GPT_5_5_MEDIUM = "gpt-5.5 (medium thinking)",
GPT_5_5_HIGH = "gpt-5.5 (high thinking)",
GPT_5_5_XHIGH = "gpt-5.5 (xhigh thinking)",
- GPT_5_2_CODEX_LOW = "gpt-5.2-codex (low thinking)",
- GPT_5_2_CODEX_MEDIUM = "gpt-5.2-codex (medium thinking)",
- GPT_5_2_CODEX_HIGH = "gpt-5.2-codex (high thinking)",
- GPT_5_2_CODEX_XHIGH = "gpt-5.2-codex (xhigh thinking)",
GPT_5_4_MINI_LOW = "gpt-5.4-mini (low thinking)",
GEMINI_3_FLASH_PREVIEW_HIGH = "gemini-3-flash-preview (high thinking)",
GEMINI_3_FLASH_PREVIEW_MINIMAL = "gemini-3-flash-preview (minimal thinking)",
@@ -48,18 +44,6 @@ export const CODE_GENERATION_MODEL_DESCRIPTIONS: {
"gpt-5.5 (xhigh thinking)": {
name: "GPT 5.5 (xhigh)",
},
- "gpt-5.2-codex (low thinking)": {
- name: "GPT 5.2 Codex (low)",
- },
- "gpt-5.2-codex (medium thinking)": {
- name: "GPT 5.2 Codex (medium)",
- },
- "gpt-5.2-codex (high thinking)": {
- name: "GPT 5.2 Codex (high)",
- },
- "gpt-5.2-codex (xhigh thinking)": {
- name: "GPT 5.2 Codex (xhigh)",
- },
"gpt-5.4-mini (low thinking)": {
name: "GPT 5.4 Mini (low)",
},