FrontierCode 1.1

Une implémentation de FrontierCode 1.1 de Cognition, un benchmark privé de codage agentique qui évalue les modèles sur l'exécution de tâches de codage d'une manière qui serait réellement fusionnée par les mainteneurs de dépôts réels.

Modèles: 75 actuel, 83 au total
Médiane: 38.8%
Plage: 0.0%–64.0%

Distribution des scores

Le plus bas: 0.0%
Médiane: 38.8%
Le plus élevé: 64.0%

Classement

Inclure les modèles obsolètes

Rang	Modèle	Score
1	Claude Fable 5.2 (Medium, Default Fallback) Anthropic	64.0%
2	Claude Fable 5.2 (Low, Default Fallback) Anthropic	60.6%
3	Claude Opus 5.5 (Medium, Default Fallback) Anthropic	57.1%
4	Claude Fable 5.2 (Max, Default Fallback) Anthropic	56.2%
5	Claude Fable 5.2 (High, Default Fallback) Anthropic	55.9%
6	Claude Fable 5.2 (Xhigh, Default Fallback) Anthropic	55.5%
7	Claude Opus 5.5 (High, Default Fallback) Anthropic	54.0%
8	Claude Opus 5.5 (Xhigh, Default Fallback) Anthropic	53.5%
9	Claude Opus 5.5 (Max, Default Fallback) Anthropic	52.2%
10	GPT-6.1 Sol (Max) OpenAI	49.1%
11	Kimi K3.1 Pro Kimi	47.1%
12	GPT-6 Astra (Max) OpenAI	47.0%
13	GPT-6.1 Sol (High) OpenAI	46.6%
14	Muse Wonder (Max) Meta	46.5%
15	GPT-6.1 Sol (Xhigh) OpenAI	46.3%
16	Claude Sonnet 5.5 (Xhigh, Default Fallback) Anthropic	46.1%
17	Claude Sonnet 5.5 (Max, Default Fallback) Anthropic	45.8%
18	Claude Haiku 5.5 (Xhigh) Anthropic	45.5%
19	MiMo-V2.6-Pro Xiaomi	45.4%
20	Muse Wonder Full (Max) Meta	44.9%
21	Claude Sonnet 5.5 (High, Default Fallback) Anthropic	44.3%
22	GPT-6.1 Terra (Medium) OpenAI	43.7%
23	DeepSeek V4.1 Flash 1017 (Max) DeepSeek	43.7%
24	Claude Haiku 5.5 (High) Anthropic	43.6%
25	Gemini 4 Argon (High) Google	43.5%
26	GPT-6.1 Sol (Medium) OpenAI	43.3%
27	Qwen3.8 Max (0902) Alibaba	43.2%
28	Claude Opus 5.5 (Low, Default Fallback) Anthropic	42.7%
29	MiMo-V2.6-Flash Xiaomi	42.1%
30	GPT-6 Luna (Max) OpenAI	41.9%
31	Claude Haiku 5.5 (Max) Anthropic	41.8%
32	GPT-6.1 Terra (Xhigh) OpenAI	41.4%
33	GPT-6.1 Terra (Non-reasoning) OpenAI	41.0%
34	GLM-5.3-Flash Z AI	41.0%
35	Claude Haiku 5.5 (Medium) Anthropic	40.4%
36	GLM-5.3 (Max) Z AI	40.1%
37	Grok 4.7 (Low) SpaceXAI	39.6%
38	DeepSeek V4 Pro 0813 (Max) DeepSeek	38.8%
39	MiniMax-M3.1-Flash MiniMax	38.6%
40	Step 5 Preview StepFun	38.4%
41	Mistral Large 4 (Medium) Mistral	37.8%
42	GPT-6.1 Terra (High) OpenAI	37.2%
43	Gemini 3.8 Flash (Medium) Google	37.1%
44	GPT-6.1 Sol (Low) OpenAI	36.7%
45	Gemini 3.8 Flash (High) Google	36.4%
46	GPT-6.1 Terra (Low) OpenAI	36.3%
47	GPT-6.1 Terra (Max) OpenAI	36.0%
48	GLM-5.3 (Low) Z AI	34.9%
49	Qwen3.8 27B (Medium) Alibaba	34.7%
50	Qwen3.8 27B (Xhigh) Alibaba	34.2%

1
Claude Fable 5.2 (Medium, Default Fallback)
Anthropic
64.0%
2
Claude Fable 5.2 (Low, Default Fallback)

Données de Artificial Analysis

Command Palette

FrontierCode 1.1

Distribution des scores

Classement