chore: merge dev into main

src/backends/claude-code/models.ts

@@ -1,4 +1,5 @@
 export const CLAUDE_CODE_MODELS = [
+	{ value: 'claude-fable-5', label: 'Claude Fable 5' },
 	{ value: 'claude-opus-4-8', label: 'Claude Opus 4.8' },
 	{ value: 'claude-opus-4-8[1m]', label: 'Claude Opus 4.8 (1M context)' },
 	{ value: 'claude-opus-4-7', label: 'Claude Opus 4.7' },

src/backends/codex/models.ts

@@ -1,4 +1,7 @@
 export const CODEX_MODELS = [
+	{ value: 'gpt-5.6-sol', label: 'GPT-5.6 Sol' },
+	{ value: 'gpt-5.6-terra', label: 'GPT-5.6 Terra' },
+	{ value: 'gpt-5.6-luna', label: 'GPT-5.6 Luna' },
 	{ value: 'gpt-5.5', label: 'GPT-5.5' },
 	{ value: 'gpt-5.4', label: 'GPT-5.4' },
 	{ value: 'gpt-5.4-mini', label: 'GPT-5.4 mini' },

src/config/rateLimits.ts

@@ -19,6 +19,13 @@ export const MODEL_RATE_LIMITS: ModelRateLimits = {
 		safetyMargin: 0.8, // Conservative - start throttling at 80%
 	},
 
+	// Claude Fable 5 (Tier 1: 50 RPM, 10K TPM — priciest Anthropic model, throttle-sensitive)
+	'anthropic:claude-fable-5': {
+		requestsPerMinute: 50,
+		tokensPerMinute: 10_000,
+		safetyMargin: 0.85,
+	},
+
 	// Claude Opus 4.8 (Tier 1: 50 RPM, 10K TPM — Opus is throttle-sensitive)
 	'anthropic:claude-opus-4-8': {
 		requestsPerMinute: 50,

src/utils/llmMetrics.ts

@@ -9,6 +9,11 @@ import type { TokenUsage } from 'llmist';
  * Prices as of January 2026.
  */
 const MODEL_PRICING: Record<string, { input: number; output: number; cachedInput?: number }> = {
+	// Anthropic Claude Fable 5 — 1M context by default (max = default), priced at 2× Opus.
+	// Key matches toPricingKey('claude-fable-5') = 'anthropic:claude-fable-5' (no trailing
+	// date to strip). cachedInput follows the 0.1× convention used by every Anthropic row.
+	'anthropic:claude-fable-5': { input: 10.0, output: 50.0, cachedInput: 1.0 },
+
 	// Anthropic Claude 4 family
 	'anthropic:claude-opus-4-8': { input: 5.0, output: 25.0, cachedInput: 0.5 },
 	'anthropic:claude-opus-4-8[1m]': { input: 5.0, output: 25.0, cachedInput: 0.5 },
@@ -28,6 +33,12 @@ const MODEL_PRICING: Record<string, { input: number; output: number; cachedInput
 	// OpenAI — Codex CLI models (per developers.openai.com/codex/pricing, 2026-05-11).
 	// Rates are public metered API prices; we display API-equivalent cost regardless of
 	// subscription plan, consistent with how Anthropic Pro/Max users see costs for claude-code.
+	// GPT-5.6 family (GA 2026-07-09, per developers.openai.com/api/docs/pricing): three
+	// Sol/Terra/Luna tiers, 1M context on all three, cached-input reads at the standard 90%
+	// discount. There is no GPT-5.6 Codex-specific model — Codex runs these same three tiers.
+	'openai:gpt-5.6-sol': { input: 5.0, output: 30.0, cachedInput: 0.5 },
+	'openai:gpt-5.6-terra': { input: 2.5, output: 15.0, cachedInput: 0.25 },
+	'openai:gpt-5.6-luna': { input: 1.0, output: 6.0, cachedInput: 0.1 },
 	'openai:gpt-5.5': { input: 5.0, output: 30.0, cachedInput: 0.5 },
 	'openai:gpt-5.4': { input: 2.5, output: 15.0, cachedInput: 0.25 },
 	// gpt-5.4-mini: rates from developers.openai.com/api/docs/models/gpt-5.4-mini (2026-05-11).

tests/unit/backends/claude-code.test.ts

@@ -295,7 +295,7 @@ describe('buildSystemPrompt', () => {
 
 describe('CLAUDE_CODE_MODELS constants', () => {
 	it('contains the expected models', () => {
-		expect(CLAUDE_CODE_MODELS).toHaveLength(10);
+		expect(CLAUDE_CODE_MODELS).toHaveLength(11);
 	});
 
 	it('includes Opus 4.8, Opus 4.7, and the 1M context variants', () => {
@@ -307,6 +307,12 @@ describe('CLAUDE_CODE_MODELS constants', () => {
 		expect(CLAUDE_CODE_MODEL_IDS).toContain('claude-opus-4-6[1m]');
 	});
 
+	it('includes Claude Fable 5 (1M context by default — no [1m] variant)', () => {
+		expect(CLAUDE_CODE_MODEL_IDS).toContain('claude-fable-5');
+		// Fable 5 is 1M-context by default, so a [1m] suffix would be redundant.
+		expect(CLAUDE_CODE_MODEL_IDS).not.toContain('claude-fable-5[1m]');
+	});
+
 	it('has value/label pairs', () => {
 		for (const m of CLAUDE_CODE_MODELS) {
 			expect(m.value).toBeTruthy();
@@ -325,6 +331,7 @@ describe('CLAUDE_CODE_MODELS constants', () => {
 
 describe('resolveClaudeModel', () => {
 	it('passes through known Claude Code model IDs', () => {
+		expect(resolveClaudeModel('claude-fable-5')).toBe('claude-fable-5');
 		expect(resolveClaudeModel('claude-opus-4-8')).toBe('claude-opus-4-8');
 		expect(resolveClaudeModel('claude-opus-4-8[1m]')).toBe('claude-opus-4-8[1m]');
 		expect(resolveClaudeModel('claude-opus-4-7')).toBe('claude-opus-4-7');

tests/unit/backends/codex.test.ts

@@ -146,6 +146,13 @@ describe('resolveCodexModel', () => {
 		expect(resolveCodexModel(`openai:${DEFAULT_CODEX_MODEL}`)).toBe(DEFAULT_CODEX_MODEL);
 	});
 
+	it('passes through the GPT-5.6 Sol/Terra/Luna tiers (bare and openai:-prefixed)', () => {
+		for (const id of ['gpt-5.6-sol', 'gpt-5.6-terra', 'gpt-5.6-luna']) {
+			expect(resolveCodexModel(id)).toBe(id);
+			expect(resolveCodexModel(`openai:${id}`)).toBe(id);
+		}
+	});
+
 	it('throws for incompatible models', () => {
 		expect(() => resolveCodexModel('openrouter:google/gemini-3-flash-preview')).toThrow(
 			'not compatible with the Codex engine',