Frontier Log

Benchmark · InactiveNo longer reported by the major labs since May 2026

LiveCodeBench, no release named

Created by UC Berkeley, MIT and Cornell, 12 Mar 2024 · Higher is better · Website

020406080100Jul 2024Oct 2024Jan 2025Apr 2025Jul 2025Oct 2025Jan 2026Apr 2026Jul 202693.5%DeepSeek-V4-Pro29.7%Gemma 3 27B83.0%MiniMax M271.1%Nemotron Nano 9B v249.2%DeepSeek-V3-032440.5%DeepSeek-V320.4%Gemma 2 27B020406080100Jul 2024Jan 2025Jul 2025Jan 2026Jul 202693.5%DeepSeek-V4-Pro29.7%Gemma 3 27B83.0%MiniMax M249.2%DeepSeek-V3-032420.4%Gemma 2 27B

Results

ModelScoreSourceSource check
Qwen3.7-Max20 May 2026 · Alibaba91.6%Qwen3.7: The Agent FrontierOur capture
CheckedA person checked this number against its source on 19 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced.
DeepSeek-V4-Flash24 Apr 2026 · DeepSeek91.6%DeepSeek V4 Preview Releasemax reasoning effortSource image
CorroboratedThe same number is printed in 1 other lab document: 1
DeepSeek-V4-Pro24 Apr 2026 · DeepSeek93.5%DeepSeek V4 Preview Releasemax reasoning effortSource image
CheckedA person checked this number against its source on 18 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced.
Mistral Small 4 119B16 Mar 2026 · Mistral63.6%Introducing Mistral Small 4high reasoning effortSource image
MiniMax M2.123 Dec 2025 · MiniMax78.0%MiniMax M2.1: Significantly Enhanced Multi-Language Programming, Built for Real-World Complex Tasksinternal run following Artificial Analysis Intelligence Index methodsSource image
MiniMax M2.123 Dec 2025 · MiniMax81.0%MiniMaxAI/MiniMax-M2.1 model cardinternal run following Artificial Analysis Intelligence Index methods; another report, not chartedOur capture
CorroboratedThe same number is printed in 1 other lab document: 1
Mistral Large 3 675B2 Dec 2025 · Mistral34.4%Introducing Mistral 3base model, before instruction tuning; no chain of thoughtSource image
CheckedA person checked this number against its source on 20 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced.
GigaChat 3 Lightning21 Nov 2025 · Sber20.31%GigaChat 3 Ultra Preview — тяжёлый open source0-shotOur capture
LongCat-Flash-Omni3 Nov 2025 · Meituan52.64%LongCat-Flash-Omni model card (Hugging Face)Our capture
CorroboratedThe same number is printed in 2 other lab documents: 1 2
MiniMax M227 Oct 2025 · MiniMax83.0%MiniMax M2.1: Significantly Enhanced Multi-Language Programming, Built for Real-World Complex Tasksinternal run following Artificial Analysis Intelligence Index methods; reported laterSource image
CorroboratedThe same number is printed in 2 other lab documents: 1 2
LongCat-Flash-Chat1 Sep 2025 · Meituan48.02%LongCat-Flash-Chat model card (Hugging Face)non-thinkingOur capture
CorroboratedThe same number is printed in 5 other lab documents, the first 3 linked: 1 2 3
DeepSeek-V3.121 Aug 2025 · DeepSeek74.8%DeepSeek-V3.1 Releasethinking mode; 13,977 output tokens on averageSource image
CheckedA person checked this number against its source on 19 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced.
Nemotron Nano 9B v218 Aug 2025 · NVIDIA71.1%NVIDIA Nemotron Nano 2 and the Nemotron Pretraining Dataset v1reasoning onSource image
CorroboratedThe same number is printed in 4 other lab documents, the first 3 linked: 1 2 3
DeepSeek-V3-032425 Mar 2025 · DeepSeek49.2%DeepSeek-V3-0324 ReleaseSource image
CheckedA person checked this number against its source on 29 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced.
CorroboratedThe same number is printed in 1 other lab document: 1
Llama Nemotron Super 49B v118 Mar 2025 · NVIDIA45.5%Build More Accurate and Efficient AI Agents with the New NVIDIA Llama Nemotron Super v1.5reported laterSource image
CorroboratedThe same number is printed in 3 other lab documents: 1 2 3
Gemma 3 27B12 Mar 2025 · Google29.7%Gemma 3 Technical Reportinstruction-tuned; 0-shot; average of 8 samples
CheckedA person checked this number against its source on 29 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced.
Gemma 3 27B12 Mar 2025 · Google39.0%Gemma 3 Technical Reportchain of thought; instruction-tuned; 0-shot; average of 8 samples; another report, not charted
CheckedA person checked this number against its source on 29 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced.
Codestral 25.0113 Jan 2025 · Mistral37.9%Codestral 25.01Our capture
DeepSeek-V326 Dec 2024 · DeepSeek40.5%Introducing DeepSeek-V3chain-of-thought; 8K output limitSource image
CheckedA person checked this number against its source on 29 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced.
CorroboratedThe same number is printed in 1 other lab document: 1
DeepSeek-V326 Dec 2024 · DeepSeek37.6%Introducing DeepSeek-V3without chain-of-thought; 8K output limit; another report, not chartedSource image
CheckedA person checked this number against its source on 29 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced.
CorroboratedThe same number is printed in 1 other lab document: 1
Gemma 2 27B27 Jun 2024 · Google20.4%Gemma 3 Technical Reportinstruction-tuned; 0-shot; average of 8 samples; reported laterOur capture
CheckedA person checked this number against its source on 20 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced.
Gemma 2 27B27 Jun 2024 · Google29.0%Gemma 3 Technical Reportchain of thought; instruction-tuned; 0-shot; average of 8 samples; reported later; another report, not charted
CheckedA person checked this number against its source on 29 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced.

All benchmarks