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 provides a complete dashboard with highly customizable domain-specific average ranks and precise efficiency metrics (average answer cost, average answer duration, P99 Latency). For further detail also check our Hugging Face Leaderboard.

Model	Score	Avg Cost ($ Cents)	Avg Latency (sec)	P99 Latency (sec)	Iterations
phi-4	3.444 (#38)	0.01 (#1)	15s (#7)	60s (#4)	205
Llama-4-scout	3.611 (#35)	0.02 (#3)	15s (#8)	60s (#5)	205
Phi-3-mini-128k-instruct	2.9 (#40)	0.02 (#4)	20s (#11)	143s (#17)	186
nova-lite-v1	3.513 (#36)	0.02 (#2)	7s (#1)	42s (#1)	205
Qwen3-30B-A3B-Instruct-2507	4.46 (#19)	0.03 (#6)	22s (#13)	174s (#26)	204
gemma-3-27b-it	4.339 (#25)	0.03 (#7)	31s (#18)	112s (#15)	204
nemotron-nano-9b-v2	3.434 (#39)	0.03 (#5)	18s (#10)	89s (#11)	194
llama-4-maverick	3.659 (#34)	0.05 (#8)	12s (#5)	65s (#7)	205
deepSeek-V3-0324	4.183 (#30)	0.07 (#10)	26s (#16)	101s (#14)	205
gemini-2.5-flash-lite	4.329 (#26)	0.07 (#9)	11s (#4)	80s (#9)	200
gpt-oss-120b	4.574 (#11)	0.07 (#11)	35s (#20)	153s (#21)	205
grok-4.1-fast-thinking	4.64 (#4)	0.07 (#12)	45s (#25)	177s (#27)	197
deepSeek-v3.2-exp	4.378 (#23)	0.08 (#14)	71s (#36)	381s (#40)	194
grok-4.1-fast	4.582 (#10)	0.08 (#13)	24s (#15)	65s (#6)	197
Deepseek-v3.1	4.377 (#24)	0.10 (#17)	29s (#17)	156s (#22)	205
grok-3-mini	4.32 (#27)	0.10 (#16)	23s (#14)	97s (#13)	204
magistral-small-2506	3.911 (#32)	0.10 (#15)	8s (#2)	56s (#3)	203
llama-3.3-nemotron-super-49b-v1.5	4.269 (#29)	0.11 (#18)	36s (#22)	166s (#25)	196
Qwen3-235B-A22B-Thinking-2507	4.585 (#9)	0.13 (#19)	74s (#37)	255s (#34)	193
GLM-4.5-Air	4.279 (#28)	0.16 (#21)	35s (#21)	144s (#20)	196
nova-pro-v1	3.476 (#37)	0.16 (#20)	8s (#3)	46s (#2)	205
Kimi-K2-Instruct	4.517 (#17)	0.21 (#22)	21s (#12)	86s (#10)	205
Llama-3.1-nemotron-ultra-253b-v1	4.163 (#31)	0.21 (#23)	36s (#23)	162s (#24)	203
deepSeek-R1-0528	4.536 (#14)	0.30 (#24)	54s (#30)	159s (#23)	198
Mistral-large-2512	4.586 (#8)	0.33 (#25)	62s (#31)	143s (#18)	175
GLM-4.5	4.556 (#13)	0.34 (#26)	51s (#28)	201s (#29)	204
Minimax-m2	4.524 (#16)	0.36 (#27)	68s (#34)	239s (#33)	193
Qwen3-next-80b-a3b-thinking	4.439 (#22)	0.40 (#28)	32s (#19)	127s (#16)	204
gemini-2.5-flash	4.475 (#18)	0.43 (#29)	17s (#9)	90s (#12)	204
claude-3.5-haiku	3.676 (#33)	0.67 (#30)	12s (#6)	73s (#8)	205
Kimi-K2-thinking	4.559 (#12)	0.80 (#31)	68s (#33)	360s (#38)	192
gpt-5-mini	4.594 (#7)	0.81 (#32)	74s (#38)	224s (#31)	196
claude-haiku-4.5	4.445 (#21)	1.95 (#33)	53s (#29)	365s (#39)	196
claude-sonnet-4.5	4.453 (#20)	2.08 (#34)	42s (#24)	283s (#35)	203
grok-4	4.535 (#15)	3.41 (#35)	70s (#35)	220s (#30)	197
gemini-3-pro-preview	4.642 (#3)	3.88 (#36)	46s (#26)	143s (#19)	194
gemini-2.5-pro	4.63 (#5)	3.95 (#37)	50s (#27)	187s (#28)	205
gpt-5	4.827 (#2)	5.43 (#38)	112s (#39)	312s (#36)	192
Claude-opus-4.5	4.6 (#6)	7.31 (#39)	66s (#32)	238s (#32)	194
gpt-5.1	4.849 (#1)	7.70 (#40)	141s (#40)	348s (#37)	195