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

Evaluate model performance across 10 specialized knowledge domains with normalized scores. Identify which models excel in specific areas—whether you need strong coding capabilities, creative writing prowess, or expertise in science and mathematics. Average scores reflect overall domain competency.

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