Request a pilot
‹ PublicAI Index

The LLM benchmark aggregator.

Qwen3.8 27B

Alibaba · 28B · open weights

Strongest in Instruction following (#16 of 57), weakest in Medical (#113 of 140). Above par in 19 of 26 scopes. Among the models it meets almost everywhere, it finishes behind Claude Fable 5 and Claude Fable 5.1 and ahead of GPT-6 Luna and O4 Mini.

§ 1 · Profile

What it is good at

Bars run from 50 — the average of the models each source lists — so right of the line is above par. The middle column is the gap to whoever leads that scope.

Agents60.8#23/268
Knowledge work64#20/178
Tool use68.5—/81✱
Coding53#48/165
Agentic coding54.8#39/157
Code generation46.5#55/77
Professional52#59/168
Legal57.1#27/151
Finance52.7#70/151
Medical44.5#113/140
Core abilities52.5#66/204
Instruction following58.4#16/57
Data analysis55.4#24/57
Language40.6#44/57
General intelligence53.9#67/204
Long context60.9—/0✱
Knowledge53.2#67/138
Academic knowledge53.9#62/123
Factuality49—/0✱
Human preference60.7#79/342
Human preference60.7#79/342
Reasoning47.6#103/178
Science58.9#32/122
Reasoning49.5#63/140
Mathematics40.6#109/140
Expert reasoning53.8—/0✱

§ 2 · Head to head

What it beats, and what beats it

The same models turn up scope after scope. Counted once: where both were placed, who finished higher. Bars run right for Qwen3.8 27B, left for the other.

Llama 4 Scout Instructwon 17 · lost 0
Mistral Medium 3.5won 17 · lost 0
GPT-4o Miniwon 18 · lost 0
GPT-4.1 Nanowon 19 · lost 0
Command Awon 16 · lost 1
Qwen3.6 27Bwon 15 · lost 1
Gemini 2.0 Flashwon 15 · lost 1
GPT-4.1 Miniwon 17 · lost 2
DeepSeek V3won 17 · lost 2
Mistral Large 3won 16 · lost 2
GPT-5.4 Nanowon 19 · lost 3
Gemini 2.5 Flashwon 16 · lost 3
GPT OSS 120Bwon 15 · lost 3
Kimi K2won 15 · lost 3
MiniMax M2.5won 13 · lost 3
Claude Haiku 4.5won 15 · lost 4
GLM-4.6won 14 · lost 4
O3 Miniwon 13 · lost 4
GPT-5.4 Miniwon 16 · lost 5
GPT-5 Nanowon 12 · lost 4
DeepSeek V3.2won 14 · lost 5
GPT-4.1won 13 · lost 5
Qwen3 235B A22B Instructwon 13 · lost 5
Nemotron 3 Ultrawon 14 · lost 6
Inklingwon 18 · lost 8
Grok 4 Fastwon 11 · lost 5
Inkling Smallwon 11 · lost 5
GLM-4.7won 11 · lost 5
Claude Sonnet 4won 13 · lost 6
Gemma 4 31Bwon 13 · lost 6
Grok 4.3won 15 · lost 7
Gemini 3.5 Flash Litewon 15 · lost 7
Mimo V2.5 Prowon 11 · lost 6
Grok 4.1 Fastwon 10 · lost 6
MiniMax M3won 16 · lost 10
DeepSeek R1won 11 · lost 7
Gemini 3.1 Flash Litewon 10 · lost 7
GPT-5 Miniwon 11 · lost 8
Qwen3.6 Pluswon 12 · lost 10
Kimi K2.6won 12 · lost 10
O4 Miniwon 10 · lost 9
GPT-6 Lunawon 10 · lost 9
Grok 3 Miniwon 8 · lost 8
GLM-5.1won 8 · lost 8
Grok 4won 9 · lost 9
GLM-5.3 Flashwon 11 · lost 11
Claude Sonnet 4.5won 9 · lost 10
GLM-5.2won 12 · lost 14
Qwen3.7 Maxwon 10 · lost 12
GPT-5won 8 · lost 10
Kimi K2.5won 8 · lost 10
GLM-5won 7 · lost 9
Gemini 2.5 Prowon 8 · lost 11
Claude Sonnet 4.6won 9 · lost 13
DeepSeek V4 Flashwon 9 · lost 14
GPT-5.1won 6 · lost 10
O3won 7 · lost 12
GPT-5.2won 8 · lost 14
Gemini 3.5 Flashwon 8 · lost 14
Gemini 3.6 Flashwon 8 · lost 14
Claude Opus 4won 6 · lost 11
Claude Sonnet 5won 9 · lost 17
Grok 4.5won 7 · lost 15
Claude Opus 4.5won 7 · lost 15
Muse Spark 1.1won 7 · lost 15
GPT-5.6 Lunawon 8 · lost 18
DeepSeek V4.1 Flashwon 6 · lost 16
Gemini 3.1 Prowon 6 · lost 16
Gemini 3.8 Flashwon 6 · lost 16
Claude Opus 4.6won 5 · lost 15
GPT-5.6 Terrawon 6 · lost 20
GPT-5.4won 5 · lost 17
DeepSeek V4 Prowon 5 · lost 18
GPT-6 Solwon 4 · lost 15
Claude Opus 4.8won 4 · lost 18
Gemini 3.7 Flashwon 4 · lost 18
GLM-5.3won 4 · lost 19
Gemini 3 Prowon 3 · lost 15
Muse Spark 1.2won 3 · lost 18
GPT-5.5won 3 · lost 19
Claude Opus 4.7won 3 · lost 19
GPT-5.6 Solwon 3 · lost 20
Claude Opus 5.5won 2 · lost 17
Grok 4.6won 2 · lost 20
Kimi K3won 2 · lost 21
Claude Opus 5won 2 · lost 21
Grok 4.7won 1 · lost 18
GPT-6 Astrawon 1 · lost 18
Qwen3.8 Maxwon 1 · lost 21
Muse Spark 1.3won 0 · lost 18
Claude Fable 5.1won 0 · lost 22
Claude Fable 5won 0 · lost 22

§ 3 · Sources

Where the numbers come from

8 publications, 36 figures. Every one links to the page it was read from.

LMArena Text ↗1 measureread 2026-09-26
LMArena Text 1438
Artificial Analysis ↗1 measureread 2026-09-26
Artificial Analysis Intelligence Index 34
GDPval-AA ↗1 measureread 2026-09-26
GDPval-AA 1409
AA-Briefcase ↗1 measureread 2026-09-26
AA-Briefcase 1400
LiveBench ↗8 measuresread 2026-09-26
LiveBench 75.3LiveBench · Reasoning 80LiveBench · Coding 75.7LiveBench · Agentic Coding 61.4LiveBench · Mathematics 86.2LiveBench · Data Analysis 76.6LiveBench · Language 74.3LiveBench · Instruction Following 72.7
SimpleBench ↗1 measureread 2026-09-26
SimpleBench 60.2%
Vals.ai ↗14 measuresread 2026-09-26
Vals · Legal Research Bench 36.06%Vals · LegalBench 82.43%Vals · Harvey Legal Agent Benchmark 11.25%Vals · Finance Agent 48.55%Vals · TaxEval 70.85%Vals · MortgageTax 64.94%Vals · MedCode 28.7%Vals · MedScribe 83.85%Vals · SWE-bench Verified 86% (Mini-SWE-agent)Vals · Vibe Code Bench 64.85% (OpenHands)Vals · Code Migration 14.16%Vals · GPQA Diamond 88.89%Vals · MMLU Pro 84.34%Vals · ProofBench 16%
tau3-Banking 48%Terminal-Bench 2.1 79.8%SciCode 44.7%Humanity's Last Exam (without tools) 33.9%GPQA Diamond 90.5%CritPt 5.4%AA-LCR 77.3%AA-Omniscience Accuracy 15.6%AA-Omniscience Non-Hallucination 69.7%

Badge

PublicAI Index badge for Qwen3.8 27B[![PublicAI Index](https://publicai.io/model-index/badge?model=qwen3-8-27b)](https://publicai.io/model-index/m/qwen3-8-27b)

✱ Placed by a one-off publication, not a board that re-ran the model. ✱✱ marks a figure the model’s own publisher printed, which counts for less again. Scores are 0–100 on the PublicAI Index scale, 50 = the average of the models each source lists. Scores belong to their publishers.

Snapshot 2026-09-26 · the full index · JSON API