AA-LCR v1.1: Long Context Reasoning

Ein anspruchsvoller Benchmark, der die Fähigkeit von Sprachmodellen misst, Informationen aus langen Dokumenten von 10k bis 100k Token (gemessen mit dem cl100k_base-Tokenizer) zu extrahieren, über sie zu schlussfolgern und sie zu synthetisieren.

Modelle: 259 aktuell, 596 insgesamt
Median: 61.3%
Bereich: 0.0%–88.3%

Bewertungsverteilung

Niedrigste: 0.0%
Median: 61.3%
Höchste: 88.3%

Bestenliste

Veraltete Modelle einschließen

Rang	Modell	Ergebnis
1	Step 5 Preview StepFun	88.3%
2	Kimi K3.1 Pro Kimi	86.3%
3	MiMo-V2.6-Pro Xiaomi	86.3%
4	Agnes 2.5 Pro Theta Sapiens AI	84.7%
5	Claude Opus 5.5 (Max, Default Fallback) Anthropic	84.7%
6	Claude Opus 5.5 (Xhigh, Default Fallback) Anthropic	84.7%
7	Claude Opus 5.5 (Medium, Default Fallback) Anthropic	84.3%
8	Gemini 3.8 Flash (Medium) Google	84.0%
9	GPT-6.1 Sol (Low) OpenAI	84.0%
10	Muse Wonder Full (Max) Meta	83.7%
11	Claude Fable 5.2 (Max, Default Fallback) Anthropic	83.3%
12	Claude Fable 5.2 (High, Default Fallback) Anthropic	83.3%
13	GPT-5.3 Codex (Xhigh) OpenAI	83.3%
14	GPT-6.1 Sol (Medium) OpenAI	83.3%
15	GPT-6 Luna (Max) OpenAI	83.3%
16	Inkling 2 Small Thinking Machines	83.3%
17	Muse Glimmer (High) Meta	83.3%
18	Solar Mini 4 Upstage	83.3%
19	Agnes 2.5 Pro Beta Sapiens AI	83.0%
20	GPT-6.1 Sol (Max) OpenAI	83.0%
21	Ling 3.1 Flash InclusionAI	83.0%
22	MiniMax-M3 MiniMax	83.0%
23	Claude Fable 5.2 (Xhigh, Default Fallback) Anthropic	82.7%
24	Claude Haiku 5.5 (Max) Anthropic	82.7%
25	Claude Opus 5.5 (High, Default Fallback) Anthropic	82.7%
26	Claude Sonnet 5.5 (Max, Default Fallback) Anthropic	82.7%
27	DeepSeek V4.1 Flash 1017 (Max) DeepSeek	82.7%
28	GPT-6.1 Terra (Max) OpenAI	82.7%
29	GPT-6.1 Sol (High) OpenAI	82.3%
30	GPT-6.1 Terra (High) OpenAI	82.3%
31	Ling 3.1 Flash 1019 InclusionAI	82.3%
32	Gemini 3.1 Pro Preview Google	82.0%
33	Qwen3.8 27B (Xhigh) Alibaba	82.0%
34	GPT-6.1 Terra (Xhigh) OpenAI	81.7%
35	Nex-N2-Pro (Based on Qwen3.5-397B-A17B) Nex AGI	81.7%
36	DeepSeek V4 Flash Vision (Max) DeepSeek	81.3%
37	Gemini 3.8 Flash (High) Google	81.3%
38	GPT-6.1 Terra (Low) OpenAI	81.0%
39	Claude Opus 5.5 (Low, Default Fallback) Anthropic	80.7%
40	Gemini 3.8 Flash (Low) Google	80.7%
41	GPT-6.1 Terra (Medium) OpenAI	80.7%
42	GPT-6 Astra (Max) OpenAI	80.7%
43	Muse Wonder (Max) Meta	80.7%
44	DeepSeek V4 Pro 0813 (Max) DeepSeek	80.3%
45	Qwen3.8 2.4T A95B Alibaba	80.3%
46	Qwen3.8 Max (0902) Alibaba	80.3%
47	Claude Fable 5.2 (Low, Default Fallback) Anthropic	80.0%
48	Claude Fable 5.2 (Medium, Default Fallback) Anthropic	80.0%
49	GLM-5.3-Flash Z AI	80.0%
50	GPT-6 Astra (High) OpenAI	80.0%

1
Step 5 Preview
StepFun
88.3%
2
Kimi K3.1 Pro

Daten von Artificial Analysis

Command Palette

AA-LCR v1.1: Long Context Reasoning

Bewertungsverteilung

Bestenliste