Skip to content

Never stop coding. Free MIT AI gateway: one endpoint, 375 providers (150+ free), 1,400+ models Kimi, Claude, GPT, Gemini, GLM, DeepSeek, MiniMax. Works with Claude Code, Codex, Cursor, OpenCode, Cline & Copilot. Quota-aware auto-fallback, RTK+Caveman compression saves 15-95% tokens, MCP/A2A, Desktop/PWA. Built by 900+ contributors

fix(claude): skip adaptive thinking defaults for unsupported models (#1563)

Cong Vu Chi31acf52d0a58
diff --git a/open-sse/executors/base.ts b/open-sse/executors/base.ts
index 38c9d62dfc..fc34faffd5 100644
--- a/open-sse/executors/base.ts
+++ b/open-sse/executors/base.ts
@@ -10,6 +10,7 @@ import {
modelSupportsContext1mBeta,
} from "../services/claudeCodeCompatible.ts";
import { getClaudeCodeCompatibleRequestDefaults } from "@/lib/providers/requestDefaults";
+import { supportsXHighEffort } from "../config/providerModels.ts";
import { remapToolNamesInRequest } from "../services/claudeCodeToolRemapper.ts";
import { obfuscateInBody } from "../services/claudeCodeObfuscation.ts";
import {
@@ -502,17 +503,19 @@ export class BaseExecutor {
};
}
- if (!tb.thinking) {
+ const supportsAdaptiveThinking = supportsXHighEffort("claude", model);
+
+ if (supportsAdaptiveThinking && !tb.thinking) {
tb.thinking = { type: "adaptive" };
}
- if (!tb.context_management) {
+ if (supportsAdaptiveThinking && !tb.context_management) {
tb.context_management = {
edits: [{ type: "clear_thinking_20251015", keep: "all" }],
};
}
- if (!tb.output_config) {
+ if (supportsAdaptiveThinking && !tb.output_config) {
tb.output_config = { effort: "high" };
}
diff --git a/tests/unit/executor-default-base.test.ts b/tests/unit/executor-default-base.test.ts
index 704f075507..d989481500 100644
--- a/tests/unit/executor-default-base.test.ts
+++ b/tests/unit/executor-default-base.test.ts
@@ -514,10 +514,13 @@ test("DefaultExecutor.execute uses CC-compatible connection defaults to append 1
credentials: {
apiKey: "cc-key",
providerSpecificData: {
- baseUrl: "https://cc.example.com/v1/messages?beta=true",
ccSessionId: "session-1",
},
},
+ clientHeaders: {
+ "x-app": "cli",
+ "user-agent": "claude-cli/2.1.116 (external, cli)",
+ },
extendedContext: false,
});
await cc.execute({
@@ -531,7 +534,6 @@ test("DefaultExecutor.execute uses CC-compatible connection defaults to append 1
credentials: {
apiKey: "cc-key",
providerSpecificData: {
- baseUrl: "https://cc.example.com/v1/messages?beta=true",
ccSessionId: "session-1",
requestDefaults: { context1m: true },
},
@@ -565,6 +567,76 @@ test("DefaultExecutor.execute uses CC-compatible connection defaults to append 1
assert.equal(calls[2].headers["anthropic-beta"], CONTEXT_1M_BETA_HEADER);
});
+test("DefaultExecutor.execute only injects adaptive thinking defaults for Claude models that support x-high effort", async () => {
+ const originalFetch = globalThis.fetch;
+ const requestBodies = [];
+
+ globalThis.fetch = async (_url, init = {}) => {
+ requestBodies.push(JSON.parse(String(init.body)));
+ return new Response(JSON.stringify({ ok: true }), {
+ status: 200,
+ headers: { "Content-Type": "application/json" },
+ });
+ };
+
+ try {
+ const claude = new DefaultExecutor("claude");
+ await claude.execute({
+ model: "claude-opus-4-7",
+ body: {
+ model: "claude-opus-4-7",
+ messages: [{ role: "user", content: "hi" }],
+ max_tokens: 1,
+ },
+ stream: false,
+ credentials: {
+ apiKey: "cc-key",
+ providerSpecificData: {
+ ccSessionId: "session-1",
+ },
+ },
+ clientHeaders: {
+ "x-app": "cli",
+ "user-agent": "claude-cli/2.1.116 (external, cli)",
+ },
+ extendedContext: false,
+ });
+
+ await claude.execute({
+ model: "claude-haiku-4-5-20251001",
+ body: {
+ model: "claude-haiku-4-5-20251001",
+ messages: [{ role: "user", content: "hi" }],
+ max_tokens: 1,
+ },
+ stream: false,
+ credentials: {
+ apiKey: "cc-key",
+ providerSpecificData: {
+ ccSessionId: "session-1",
+ },
+ },
+ clientHeaders: {
+ "x-app": "cli",
+ "user-agent": "claude-cli/2.1.116 (external, cli)",
+ },
+ extendedContext: false,
+ });
+ } finally {
+ globalThis.fetch = originalFetch;
+ }
+
+ assert.deepEqual((requestBodies[0] as any).thinking, { type: "adaptive" });
+ assert.deepEqual((requestBodies[0] as any).context_management, {
+ edits: [{ type: "clear_thinking_20251015", keep: "all" }],
+ });
+ assert.deepEqual((requestBodies[0] as any).output_config, { effort: "high" });
+
+ assert.equal((requestBodies[1] as any).thinking, undefined);
+ assert.equal((requestBodies[1] as any).context_management, undefined);
+ assert.equal((requestBodies[1] as any).output_config, undefined);
+});
+
test("DefaultExecutor.transformRequest is a passthrough and preserves model ids with slashes", () => {
const executor = new DefaultExecutor("openai");
const body = { model: "zai-org/GLM-5-FP8", messages: [{ role: "user", content: "hi" }] };