Benchmark Almanac

AutoBench Agronomy LLM Benchmark

GPT-5.1 4.85 leads since December 2025
Not in model cards

Models answer questions from three types of farm operators and an agricultural researcher. Other models rank the answers.

The December 2025 agronomy leaderboard reports scores and costs for 40 models. AutoBench reports 20 judges and more than 200 generated questions.

Made by
AutoBench and EVJA
Released
December 2025
Kind
Tasks from real jobs
Level
University
Scored
Graded by a model
Answers
Open-ended
Measures
Preference
Language
English
In and out
Text → Text
Interaction
Single turn
Items from
Generated
Refresh
One fixed set
Run by
A host
Occupations
Life, physical and social science, Farming, fishing and forestry
Strongest model tested
Gemini 3 Pro #56 on the capability index
Last model added
November 2025
Scores refreshed
November 20th, 2026
39
Entries on its own leaderboard, 34 models we track

GPT-5.1 leads with 4.85 among models released by November 2025 GPT-5.1 leads with 4.85 among models released by November 2025

Best score per entry from the benchmark's own leaderboard, read on November 20th, 2026. Where a model ran with several settings or harnesses, we keep its best. Every run is in the next chart. Names in gray aren't in our model list: fine-tunes, agent systems, products and older models. They don't count toward leaders. This view shows only models we track, each at its best.

The newest model with a score here came out in November 2025, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:

  1. 1 GPT-5.1 4.85
  2. 2 GPT-5 4.83
  3. 3 Gemini 3 Pro 4.64
  4. 4 Grok 4.1 Fast thinking 4.64
  5. 5 Gemini 2.5 Pro Jun 2025 4.63
  6. 6 Claude Opus 4.5 4.60
  7. 7 GPT-5 mini 4.59
  8. 8 Mistral-large-2512 4.59
  9. 9 Qwen3-235B-A22B-Thinking Jul 2025 4.58
  10. 10 gpt-oss-120b 4.57
  11. 11 Kimi K2 Thinking 4.56
  12. 12 GLM-4.5 4.56
  13. 13 DeepSeek-R1 May 2025 4.54
  14. 14 Grok 4 4.54
  15. 15 MiniMax-M2 4.52
  16. 16 Kimi K2 Jul 2025 4.52
  17. 17 Gemini 2.5 Flash Sep 2025 4.47
  18. 18 Qwen3-30B-A3B-Instruct Jul 2025 4.46
  19. 19 Claude Sonnet 4.5 4.45
  20. 20 Claude Haiku 4.5 4.45
  21. 21 Qwen3-Next-80B-A3B thinking 4.44
  22. 22 DeepSeek-V3.2-Exp 4.38
  23. 23 DeepSeek-V3.1 4.38
  24. 24 Gemma 3 27B 4.34
  25. 25 Gemini 2.5 Flash-Lite Sep 2025 4.33
  26. 26 Grok-3 mini 4.32
  27. 27 GLM-4.5-Air 4.28
  28. 28 Llama-3.3-Nemotron-Super-49B-v1.5 4.27
  29. 29 DeepSeek-V3 Mar 2025 4.18
  30. 30 Llama-3.1-Nemotron-Ultra-253B-v1 4.16
  31. 31 Magistral-small-2506 3.91
  32. 32 Claude 3.5 Haiku 3.68
  33. 33 Llama 4 Maverick 3.66
  34. 34 Llama 4 Scout 3.61
  35. 35 Nova-lite-v1 3.51
  36. 36 Nova-pro-v1 3.48
  37. 37 Phi-4 3.44
  38. 38 Nemotron-nano-9b-v2 3.43
  39. 39 Phi-3-mini-128k-instruct 2.90

Source: Official AutoBench Agronomy LLM Benchmark leaderboard

Source: Official AutoBench Agronomy LLM Benchmark leaderboard

#EntryScoreSettingDateSource
1 GPT-5.1 4.85 Nov 2025 Official leaderboard
2 GPT-5 4.83 Aug 2025 Official leaderboard
3 Gemini 3 Pro 4.64 Nov 2025 Official leaderboard
4 Grok 4.1 Fast 4.64 thinking Nov 2025 Official leaderboard
5 Gemini 2.5 Pro Jun 2025 4.63 Jun 2025 Official leaderboard
6 Claude Opus 4.5 4.60 Nov 2025 Official leaderboard
7 GPT-5 mini 4.59 Aug 2025 Official leaderboard
8 Mistral-large-2512 4.59 Official leaderboard
9 Qwen3-235B-A22B-Thinking Jul 2025 4.58 Jul 2025 Official leaderboard
10 gpt-oss-120b 4.57 Aug 2025 Official leaderboard
11 Kimi K2 Thinking 4.56 Nov 2025 Official leaderboard
12 GLM-4.5 4.56 Aug 2025 Official leaderboard
13 DeepSeek-R1 May 2025 4.54 May 2025 Official leaderboard
14 Grok 4 4.54 Jul 2025 Official leaderboard
15 MiniMax-M2 4.52 Oct 2025 Official leaderboard
16 Kimi K2 Jul 2025 4.52 Jul 2025 Official leaderboard
17 Gemini 2.5 Flash Sep 2025 4.47 Sep 2025 Official leaderboard
18 Qwen3-30B-A3B-Instruct Jul 2025 4.46 Jul 2025 Official leaderboard
19 Claude Sonnet 4.5 4.45 Sep 2025 Official leaderboard
20 Claude Haiku 4.5 4.45 Oct 2025 Official leaderboard
21 Qwen3-Next-80B-A3B 4.44 thinking Sep 2025 Official leaderboard
22 DeepSeek-V3.2-Exp 4.38 Sep 2025 Official leaderboard
23 DeepSeek-V3.1 4.38 Aug 2025 Official leaderboard
24 Gemma 3 27B 4.34 Mar 2025 Official leaderboard
25 Gemini 2.5 Flash-Lite Sep 2025 4.33 Sep 2025 Official leaderboard
26 Grok-3 mini 4.32 Apr 2025 Official leaderboard
27 GLM-4.5-Air 4.28 Jul 2025 Official leaderboard
28 Llama-3.3-Nemotron-Super-49B-v1.5 4.27 Jul 2025 Official leaderboard
29 DeepSeek-V3 Mar 2025 4.18 Mar 2025 Official leaderboard
30 Llama-3.1-Nemotron-Ultra-253B-v1 4.16 Apr 2025 Official leaderboard
31 Magistral-small-2506 3.91 Official leaderboard
32 Claude 3.5 Haiku 3.68 Oct 2024 Official leaderboard
33 Llama 4 Maverick 3.66 Apr 2025 Official leaderboard
34 Llama 4 Scout 3.61 Apr 2025 Official leaderboard
35 Nova-lite-v1 3.51 Official leaderboard
36 Nova-pro-v1 3.48 Official leaderboard
37 Phi-4 3.44 Dec 2024 Official leaderboard
38 Nemotron-nano-9b-v2 3.43 Official leaderboard
39 Phi-3-mini-128k-instruct 2.90 Apr 2024 Official leaderboard
#ModelScoreSettingDateSource
1 GPT-5.1 4.85 Nov 2025 Official leaderboard
2 GPT-5 4.83 Aug 2025 Official leaderboard
3 Gemini 3 Pro 4.64 Nov 2025 Official leaderboard
4 Grok 4.1 Fast 4.64 thinking Nov 2025 Official leaderboard
5 Gemini 2.5 Pro Jun 2025 4.63 Jun 2025 Official leaderboard
6 Claude Opus 4.5 4.60 Nov 2025 Official leaderboard
7 GPT-5 mini 4.59 Aug 2025 Official leaderboard
8 Qwen3-235B-A22B-Thinking Jul 2025 4.58 Jul 2025 Official leaderboard
9 gpt-oss-120b 4.57 Aug 2025 Official leaderboard
10 Kimi K2 Thinking 4.56 Nov 2025 Official leaderboard
11 GLM-4.5 4.56 Aug 2025 Official leaderboard
12 DeepSeek-R1 May 2025 4.54 May 2025 Official leaderboard
13 Grok 4 4.54 Jul 2025 Official leaderboard
14 MiniMax-M2 4.52 Oct 2025 Official leaderboard
15 Kimi K2 Jul 2025 4.52 Jul 2025 Official leaderboard
16 Gemini 2.5 Flash Sep 2025 4.47 Sep 2025 Official leaderboard
17 Qwen3-30B-A3B-Instruct Jul 2025 4.46 Jul 2025 Official leaderboard
18 Claude Sonnet 4.5 4.45 Sep 2025 Official leaderboard
19 Claude Haiku 4.5 4.45 Oct 2025 Official leaderboard
20 Qwen3-Next-80B-A3B 4.44 thinking Sep 2025 Official leaderboard
21 DeepSeek-V3.2-Exp 4.38 Sep 2025 Official leaderboard
22 DeepSeek-V3.1 4.38 Aug 2025 Official leaderboard
23 Gemma 3 27B 4.34 Mar 2025 Official leaderboard
24 Gemini 2.5 Flash-Lite Sep 2025 4.33 Sep 2025 Official leaderboard
25 Grok-3 mini 4.32 Apr 2025 Official leaderboard
26 GLM-4.5-Air 4.28 Jul 2025 Official leaderboard
27 Llama-3.3-Nemotron-Super-49B-v1.5 4.27 Jul 2025 Official leaderboard
28 DeepSeek-V3 Mar 2025 4.18 Mar 2025 Official leaderboard
29 Llama-3.1-Nemotron-Ultra-253B-v1 4.16 Apr 2025 Official leaderboard
30 Claude 3.5 Haiku 3.68 Oct 2024 Official leaderboard
31 Llama 4 Maverick 3.66 Apr 2025 Official leaderboard
32 Llama 4 Scout 3.61 Apr 2025 Official leaderboard
33 Phi-4 3.44 Dec 2024 Official leaderboard
34 Phi-3-mini-128k-instruct 2.90 Apr 2024 Official leaderboard

The best score, 4.85, costs $0.08 a task

Runs on the cost frontier score higher than any run that's cheaper. Grok 4.1 Fast at thinking comes within 0.2 of the best for $0.0007 a task, 110 times cheaper. The costs are the leaderboard's own estimates, so compare them only within this benchmark.

Microsoft Meta Qwen OpenAI Google DeepMind xAI Cost frontier One model at each effort level
0.00 1.00 2.00 3.00 4.00 5.00 $0.0001 $0.001 $0.01 $0.10 Cost per task, log scale GPT-5.1 GPT-5 Gemini 3 Pro Grok 4.1 Fast, thinking Llama 4 Scout 0.00 1.00 2.00 3.00 4.00 5.00 $0.0001 $0.001 $0.01 $0.10 Cost per task, log scale GPT-5.1 Grok 4.1 Fast, thinking

Source: Official AutoBench Agronomy LLM Benchmark leaderboard

ModelEffortHarnessScoreCost per taskOutput tokens
GPT-5.1 4.85 $0.08
GPT-5 4.83 $0.05
Gemini 3 Pro 4.64 $0.04
Grok 4.1 Fast thinking 4.64 $0.0007
Gemini 2.5 Pro Jun 2025 4.63 $0.04
Claude Opus 4.5 4.60 $0.07
GPT-5 mini 4.59 $0.008
Qwen3-235B-A22B-Thinking Jul 2025 4.58 $0.001
Grok 4.1 Fast 4.58 $0.0008
gpt-oss-120b 4.57 $0.0007
Kimi K2 Thinking 4.56 $0.008
GLM-4.5 4.56 $0.003
DeepSeek-R1 May 2025 4.54 $0.003
Grok 4 4.54 $0.03
MiniMax-M2 4.52 $0.004
Kimi K2 Jul 2025 4.52 $0.002
Gemini 2.5 Flash Sep 2025 4.47 $0.004
Qwen3-30B-A3B-Instruct Jul 2025 4.46 $0.0003
Claude Sonnet 4.5 4.45 $0.02
Claude Haiku 4.5 4.45 $0.02
Qwen3-Next-80B-A3B thinking 4.44 $0.004
DeepSeek-V3.2-Exp 4.38 $0.0008
DeepSeek-V3.1 4.38 $0.001
Gemma 3 27B 4.34 $0.0003
Gemini 2.5 Flash-Lite Sep 2025 4.33 $0.0007
Grok-3 mini 4.32 $0.001
GLM-4.5-Air 4.28 $0.002
Llama-3.3-Nemotron-Super-49B-v1.5 4.27 $0.001
DeepSeek-V3 Mar 2025 4.18 $0.0007
Llama-3.1-Nemotron-Ultra-253B-v1 4.16 $0.002
Claude 3.5 Haiku 3.68 $0.007
Llama 4 Maverick 3.66 $0.0005
Llama 4 Scout 3.61 $0.0002
Phi-4 3.44 $0.0001
Phi-3-mini-128k-instruct 2.90 $0.0002

The best score, 4.85, costs $0.08 a task

Runs on the cost frontier score higher than any run that's cheaper. Grok 4.1 Fast at thinking comes within 0.2 of the best for $0.0007 a task, 110 times cheaper. The costs are the leaderboard's own estimates, so compare them only within this benchmark.

Microsoft Meta Qwen OpenAI Google DeepMind xAI Cost frontier One model at each effort level
0.00 1.00 2.00 3.00 4.00 5.00 $0.0001 $0.001 $0.01 $0.10 Cost per task, log scale GPT-5.1 GPT-5 Gemini 3 Pro Grok 4.1 Fast, thinking Llama 4 Scout 0.00 1.00 2.00 3.00 4.00 5.00 $0.0001 $0.001 $0.01 $0.10 Cost per task, log scale GPT-5.1 Grok 4.1 Fast, thinking

Source: Official AutoBench Agronomy LLM Benchmark leaderboard

ModelEffortHarnessScoreCost per taskOutput tokens
GPT-5.1 4.85 $0.08
GPT-5 4.83 $0.05
Gemini 3 Pro 4.64 $0.04
Grok 4.1 Fast thinking 4.64 $0.0007
Gemini 2.5 Pro Jun 2025 4.63 $0.04
Claude Opus 4.5 4.60 $0.07
GPT-5 mini 4.59 $0.008
Qwen3-235B-A22B-Thinking Jul 2025 4.58 $0.001
Grok 4.1 Fast 4.58 $0.0008
gpt-oss-120b 4.57 $0.0007
Kimi K2 Thinking 4.56 $0.008
GLM-4.5 4.56 $0.003
DeepSeek-R1 May 2025 4.54 $0.003
Grok 4 4.54 $0.03
MiniMax-M2 4.52 $0.004
Kimi K2 Jul 2025 4.52 $0.002
Gemini 2.5 Flash Sep 2025 4.47 $0.004
Qwen3-30B-A3B-Instruct Jul 2025 4.46 $0.0003
Claude Sonnet 4.5 4.45 $0.02
Claude Haiku 4.5 4.45 $0.02
Qwen3-Next-80B-A3B thinking 4.44 $0.004
DeepSeek-V3.2-Exp 4.38 $0.0008
DeepSeek-V3.1 4.38 $0.001
Gemma 3 27B 4.34 $0.0003
Gemini 2.5 Flash-Lite Sep 2025 4.33 $0.0007
Grok-3 mini 4.32 $0.001
GLM-4.5-Air 4.28 $0.002
Llama-3.3-Nemotron-Super-49B-v1.5 4.27 $0.001
DeepSeek-V3 Mar 2025 4.18 $0.0007
Llama-3.1-Nemotron-Ultra-253B-v1 4.16 $0.002
Claude 3.5 Haiku 3.68 $0.007
Llama 4 Maverick 3.66 $0.0005
Llama 4 Scout 3.61 $0.0002
Phi-4 3.44 $0.0001
Phi-3-mini-128k-instruct 2.90 $0.0002

Number one over time

Lab claims count from the day the card came out, independent runs from the model’s release.

OpenAI
Own leaderboard Not tested since Dec 2025
GPT-5.1 4.85

Source: Official AutoBench Agronomy LLM Benchmark leaderboard

Scores over time

All the scores we have, one dot each, so a model tested at several settings appears more than once.

Plot with other benchmarks
On the benchmark’s own leaderboard
2.50 3.00 3.50 4.00 4.50 5.00 2024 2025 2026 2.50 3.00 3.50 4.00 4.50 5.00 2024 2025 2026

Source: Official AutoBench Agronomy LLM Benchmark leaderboard

No lab model card reports it yet

We read the headline results of 540 launch posts from 57 labs, and none of them list this benchmark.

More in Economic work

All 210 in Economic work
  • JobBench In use 130 tasks from 35 occupations workers most want delegated. Agent works through messy reference files, graded by fact-anchored binary rubrics. GLM-5.3 61.4%
  • GDPval In use Real work tasks from 44 occupations with reference files. Experts blindly compare model deliverables with expert work. Metric is win-or-tie rate. Seed2.1 Pro 87.9%
  • GDPval-AA v2.1 In use Agents with shell and web access complete 220 GDPval gold tasks from 44 occupations. Blind pairwise comparisons give Elo ratings. Claude Opus 5.5 1,846
  • AA-Briefcase v1.1 In use Version 1.1 of Artificial Analysis's multi-hour office work evaluation, with Elo ratings fitted by Crowd-BT. Claude Opus 5.5 1,822
  • Agents' Last Exam v1.1 In use Agents complete 122 public long-horizon professional workflow tasks across 55 sub-industries. Deterministic graders check outcomes against hidden references. GPT-6.2 Astra 47.5%
  • CoWorkBench In use Agent completes long-horizon productivity tasks across computer science, finance, law and medicine. Qwen in-house scoring (method unpublished). Claude Fable 5 75.9%

Data audit

7 of 9 checks pass, 2 warnings. Look first at the Task access check. We don’t know whether the tasks are public. Every benchmark

Catalog record 1 warning 1 pass
  • Warning Task access We don’t know whether the tasks are public.
  • Pass Confidence Checked against a primary source.
Scores 1 warning 2 pass 2 notes
  • Pass Independent runs Scores from the official leaderboard (39 entries).
  • Pass Models scored 34 models we track, plus 5 other entries on its leaderboard.
  • Warning Newest score The newest came 12 months ago, in November 2025, for Claude Opus 4.5.
  • Note Frontier models None of the top 10 has a score. The strongest tested is Gemini 3 Pro (#56).
  • Note Headroom Scored as a plain number, with no fixed ceiling.
Leaderboard import 3 pass
  • Pass Importer We read the board on huggingface.co: 35 rows for models we track and 5 other entries at the last read.
  • Pass Last read November 20th, 2026.
  • Pass Read worked The last read worked.
Consistency 1 pass 1 note
  • Note Lab claims No model has both a lab number and an independent run.
  • Pass Known issues No published audit found problems.

How we check