AutoBench Agronomy LLM Benchmark - December 2025

The first AutoBench run for the Agronomy domain with models Gemini 3 Pro, Gpt 5.1, Grok 4.1, Opus 4.5 and more

Past

Date

December 10, 2025

Version

2025-12-10

Models

New Models

Run data

Overall Comparison Costs Latency P99 Latency Domain

AutoBench features 87.08% correlation with Artificial Analysis Intelligence Index, 77.16% with LMArena, and 80.68% with MMLU. Models sorted by AutoBench score.

Model	AutoBench
Phi 3 mini 128k instruct	2.9 (#40)
Nemotron nano 9b v2	3.43 (#39)
Phi 4	3.44 (#38)
Nova pro v1	3.48 (#37)
Nova lite v1	3.51 (#36)
Llama 4 scout	3.61 (#35)
Llama 4 Maverick	3.66 (#34)
Claude 3.5 haiku	3.68 (#33)
Magistral small 2506	3.91 (#32)
Llama 3.1 nemotron ultra 253b v1	4.16 (#31)
DeepSeek V3 0324	4.18 (#30)
Llama 3.3 nemotron super 49b v1.5	4.27 (#29)
GLM 4.5 Air	4.28 (#28)
Grok 3 mini	4.32 (#27)
Gemini 2.5 flash lite	4.33 (#26)
Gemma 3 27b it	4.34 (#25)
Deepseek v3.1	4.38 (#24)
Deepseek v3.2 exp	4.38 (#23)
Qwen3 next 80b a3b thinking	4.44 (#22)
Claude haiku 4.5	4.45 (#21)
Claude sonnet 4.5	4.45 (#20)
Qwen3 30b a3b instruct 2507	4.46 (#19)
Gemini 2.5 flash	4.47 (#18)
Kimi K2 Instruct	4.52 (#17)
Minimax m2	4.52 (#16)
Grok 4	4.54 (#15)
DeepSeek R1 0528	4.54 (#14)
GLM 4.5	4.56 (#13)
Kimi k2 thinking	4.56 (#12)
Gpt oss 120b	4.57 (#11)
Grok 4.1 fast	4.58 (#10)
Qwen3 235B A22B Thinking 2507	4.58 (#9)
Mistral large 2512	4.59 (#8)
Gpt 5 mini	4.59 (#7)
Claude opus 4.5	4.6 (#6)
Gemini 2.5 pro	4.63 (#5)
Grok 4.1 fast thinking	4.64 (#4)
Gemini 3 pro preview	4.64 (#3)
Gpt 5	4.83 (#2)
Gpt 5.1	4.85 (#1)