Multilingual

SWE Multilingual leaderboard

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

As of 2026-10-07, the highest measured score on SWE Multilingual is 93.9% by Claude Opus 5.5. 9 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	93.9%	measured
2	Claude Mythos 5	92.2%	measured
3	Claude Sonnet 5.5	90.3%	measured
4	Claude Opus 5	89.5%	measured
5	Claude Fable 5.1	89.1%	measured
6	Claude Opus 4.8	84.4%	measured
7	Hy4 preview	82.9%	measured
8	Qwen3.8-Flash-Next	81.0%	measured
9	Qwen3.8-Omni-Flash	80.5%	measured
10	Composer 2.5	79.8%	measured
11	Ornith-1.5-397B	79.6%	measured
12	Ornith-1.0-397B	78.9%	measured
13	Laguna S 2.1	78.5%	measured
14	Claude Sonnet 5	78.3%	measured
15	Qwen3.7 Max	78.3%	measured
16	Grok 4.5	78.0%	measured
17	Beam	78.0%	measured
18	SWE-1.7	77.8%	measured
19	Claude Opus 4.5	77.5%	measured
20	Qwen3.5 397B	76.8%	estimated ± 2.1 pp, low confidence
21	Kimi K2.6	76.7%	measured
22	MiniMax M2.7	76.5%	measured
23	DeepSeek V4 Pro 0813	76.2%	measured
24	Qwen3.7 Plus	75.8%	measured
25	dots3-note Preview	75.7%	measured
26	Qwen3.6 Plus	73.8%	measured
27	Composer 2	73.7%	measured
28	DeepSeek V4 Flash 0731	73.3%	measured
29	GLM-5	73.3%	measured
30	Kimi K2.5	73.0%	measured
31	Ling 3.0 Flash	72.4%	measured
32	Ornith-1.5-35B-A3B	71.4%	measured
33	Qwen3.6-27B	71.3%	measured
34	Qwen3.5-122B-A10B	70.5%	estimated ± 2.6 pp, low confidence
35	Qwen3.5-27B	70.5%	estimated ± 2.6 pp, low confidence
36	Ornith-1.0-35B	69.3%	measured
37	Qwen3.5-35B-A3B	68.6%	estimated ± 2.6 pp, low confidence
38	Nemotron 3 Ultra	67.7%	measured
39	Qwen3.6-35B-A3B	67.2%	measured
40	Qwen3 235B 2507	66.2%	estimated ± 2.6 pp, low confidence
41	Laguna M.1	63.1%	measured
42	Laguna XS 2.1	63.1%	measured
43	LongCat-Flash-Lite-Sparse	59.3%	measured
44	Laguna XS.2	57.7%	measured
45	GPT-4.1	56.7%	estimated ± 2.6 pp, low confidence
46	Ornith-1.5-9B	54.4%	measured
47	DeepSeek V3 0324	53.9%	estimated ± 2.6 pp, low confidence
48	Ornith-1.0-9B	52.0%	measured
49	GPT-4o	43.0%	estimated ± 2.6 pp, low confidence
50	Granite 4.2 30B	41.9%	measured
51	Nemotron 3.5 Lightning 30B A3B NVFP4	36.5%	measured
52	Phi-4	32.0%	estimated ± 2.6 pp, low confidence
53	Granite 4.2 8B	30.8%	measured
54	LLaDA2.2-flash	25.0%	measured

SWE Multilingual leaderboard

All leaderboards

Agentic · terminal

Agentic · tools

Coding

Math

Knowledge & reasoning

Instruction following

Multilingual

Vision & documents

Long context

General