Multilingual

INCLUDE leaderboard

Name: benchgap: LLM benchmark scores, measured and estimated
Creator: benchgap

As of 2026-10-10, the highest measured score on INCLUDE is 89.8% by Claude Opus 5. 51 more models have estimated scores, calibrated from the benchmarks they were measured on.

Measured scores: benchlm.ai.

#	Model	Score	Source
1	Claude Opus 5.5	90.3%	estimated ± 2.0 pp, low confidence
2	Claude Mythos 5	89.8%	estimated ± 2.0 pp, low confidence
3	Claude Opus 5	89.8%	measured
4	Claude Sonnet 5.5	89.3%	estimated ± 2.0 pp, low confidence
5	Claude Fable 5.1	89.0%	estimated ± 2.0 pp, medium confidence
6	Claude Opus 4.8	87.6%	measured
7	Hy4 preview	87.2%	estimated ± 2.0 pp, medium confidence
8	Qwen3.8-Flash-Next	86.7%	estimated ± 2.0 pp, medium confidence
9	Qwen3.8-Omni-Flash	86.6%	estimated ± 2.0 pp, medium confidence
10	Composer 2.5	86.4%	estimated ± 2.0 pp, medium confidence
11	Ornith-1.5-397B	86.3%	estimated ± 2.0 pp, medium confidence
12	Claude Opus 4.5	86.2%	measured
13	Qwen3.7 Max	86.2%	measured
14	Ornith-1.0-397B	86.1%	estimated ± 2.0 pp, medium confidence
15	Laguna S 2.1	86.0%	estimated ± 2.0 pp, medium confidence
16	Claude Sonnet 5	86.0%	estimated ± 2.0 pp, medium confidence
17	Grok 4.5	85.9%	estimated ± 2.0 pp, medium confidence
18	Beam	85.9%	estimated ± 2.0 pp, medium confidence
19	SWE-1.7	85.8%	estimated ± 2.0 pp, medium confidence
20	Qwen3.5 397B	85.6%	measured
21	Kimi K2.6	85.5%	estimated ± 2.0 pp, medium confidence
22	MiniMax M2.7	85.5%	estimated ± 2.0 pp, medium confidence
23	DeepSeek V4 Pro 0813	85.4%	estimated ± 2.0 pp, medium confidence
24	K-EXAONE 2.0	85.3%	estimated ± 1.8 pp, low confidence
25	dots3-note Preview	85.2%	estimated ± 2.0 pp, medium confidence
26	Composer 2	84.7%	estimated ± 2.0 pp, medium confidence
27	Qwen3.6 Plus	84.7%	estimated ± 1.4 pp, low confidence
28	DeepSeek V4 Flash 0731	84.6%	estimated ± 2.0 pp, medium confidence
29	Ling 3.0 Flash	84.3%	estimated ± 2.0 pp, medium confidence
30	MiMo-V2-Flash	84.1%	estimated ± 2.0 pp, medium confidence
31	Ornith-1.5-35B-A3B	84.0%	estimated ± 2.0 pp, medium confidence
32	Qwen3.6-27B	84.0%	estimated ± 2.0 pp, medium confidence
33	GLM-5	83.8%	estimated ± 1.4 pp, low confidence
34	Nemotron 3 Ultra	83.7%	estimated ± 1.4 pp, low confidence
35	Ornith-1.0-35B	83.5%	estimated ± 2.0 pp, medium confidence
36	Kimi K2.5	83.3%	measured
37	Qwen3.5-122B-A10B	83.2%	estimated ± 1.4 pp, low confidence
38	Qwen3.5-27B	83.2%	estimated ± 1.4 pp, low confidence
39	Qwen3.7 Plus	83.0%	measured
40	Qwen3.6-35B-A3B	82.9%	estimated ± 2.0 pp, low confidence
41	Qwen3 235B 2507 (Reasoning)	82.5%	estimated ± 1.4 pp, low confidence
42	Qwen3.5-35B-A3B	82.5%	estimated ± 1.4 pp, low confidence
43	Laguna M.1	81.7%	estimated ± 2.0 pp, low confidence
44	Laguna XS 2.1	81.7%	estimated ± 2.0 pp, low confidence
45	Qwen3 235B 2507	81.6%	estimated ± 1.4 pp, low confidence
46	LongCat-Flash-Lite-Sparse	80.7%	estimated ± 2.0 pp, low confidence
47	Laguna XS.2	80.2%	estimated ± 2.0 pp, low confidence
48	Ornith-1.5-9B	79.3%	estimated ± 2.0 pp, low confidence
49	Ornith-1.0-9B	78.6%	estimated ± 2.0 pp, low confidence
50	GPT-4.1	77.4%	estimated ± 1.4 pp, low confidence
51	Kimi K2	77.3%	estimated ± 2.0 pp, low confidence
52	DeepSeek V3 0324	76.0%	estimated ± 1.4 pp, low confidence
53	Granite 4.2 30B	75.8%	estimated ± 2.0 pp, low confidence
54	Nemotron 3.5 Lightning 30B A3B NVFP4	74.3%	estimated ± 2.0 pp, low confidence
55	Granite 4.2 8B	72.7%	estimated ± 2.0 pp, low confidence
56	LLaDA2.2-flash	71.1%	estimated ± 2.0 pp, low confidence
57	GPT-4o	69.4%	estimated ± 1.4 pp, low confidence
58	Phi-4	60.6%	estimated ± 1.4 pp, low confidence

INCLUDE leaderboard

All leaderboards

Agentic · terminal

Agentic · tools

Coding

Math

Knowledge & reasoning

Instruction following

Multilingual

Vision & documents

Long context

General