Benchmark Almanac

GeoNatureAgent Benchmark

Claude Sonnet 4 60.8% leads since April 2026
Not in model cards

Agents use a geospatial API to analyze environmental indicators in Spain and Portugal. The grader checks tool use and answers against task requirements.

The authors rank nine models on 93 tasks and on a combined 103 tasks. Axis Spatial separately tests a subset with Python and revised grading.

Released
April 2026
Kind
Function calling
Level
University
Scored
Checked automatically
Answers
Agent task
Tasks
All public
Measures
Capability
Language
English
In and out
Text, Maps and geodata → Text, Actions
Interaction
Agentic
Environment
APIs and tools
Items from
Real work and use
Refresh
One fixed set
Run by
Anyone
In views
Agents
Strongest model tested
GLM-5 #89 on the capability index
Last model added
February 2026
Scores refreshed
November 20th, 2026
9
Entries on its own leaderboard, 8 models we track

Claude Sonnet 4 leads with 60.8% among models released by February 2026 Claude Sonnet 4 leads with 60.8% among models released by February 2026

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 February 2026, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:

  1. 1 Claude Sonnet 4 Anthropic API 60.8%
  2. 2 DeepSeek-V3.2 Vertex AI MaaS 56.3%
  3. 3 GLM-5 Vertex AI MaaS 50.2%
  4. 4 Gemini 2.5 Pro Jun 2025, Vertex AI native 48%
  5. 5 GPT-4o Mar 2025, OpenRouter 41.6%
  6. 6 Qwen3-235B Vertex AI MaaS 41.2%
  7. 7 gpt-oss-120b Vertex AI MaaS 34.1%
  8. 8 Llama 4 Scout Vertex AI MaaS 26.9%
  9. 9 Gemma 3 27B OpenRouter 15.8%

Source: Official GeoNatureAgent Benchmark leaderboard

#EntryScore ±SettingDateSource
1 Claude Sonnet 4 60.8% ±0.8 Anthropic API May 2025 Official leaderboard
2 DeepSeek-V3.2 56.3% ±3.1 Vertex AI MaaS Dec 2025 Official leaderboard
3 GLM-5 50.2% ±2.2 Vertex AI MaaS Feb 2026 Official leaderboard
4 Gemini 2.5 Pro Jun 2025 48% ±3.3 Vertex AI native Jun 2025 Official leaderboard
5 GPT-4o Mar 2025 41.6% ±2.7 OpenRouter Mar 2025 Official leaderboard
6 Qwen3-235B 41.2% ±4.3 Vertex AI MaaS Official leaderboard
7 gpt-oss-120b 34.1% ±1.2 Vertex AI MaaS Aug 2025 Official leaderboard
8 Llama 4 Scout 26.9% ±2.1 Vertex AI MaaS Apr 2025 Official leaderboard
9 Gemma 3 27B 15.8% ±1.6 OpenRouter Mar 2025 Official leaderboard
#ModelScore ±SettingDateSource
1 Claude Sonnet 4 60.8% ±0.8 Anthropic API May 2025 Official leaderboard
2 DeepSeek-V3.2 56.3% ±3.1 Vertex AI MaaS Dec 2025 Official leaderboard
3 GLM-5 50.2% ±2.2 Vertex AI MaaS Feb 2026 Official leaderboard
4 Gemini 2.5 Pro Jun 2025 48% ±3.3 Vertex AI native Jun 2025 Official leaderboard
5 GPT-4o Mar 2025 41.6% ±2.7 OpenRouter Mar 2025 Official leaderboard
6 gpt-oss-120b 34.1% ±1.2 Vertex AI MaaS Aug 2025 Official leaderboard
7 Llama 4 Scout 26.9% ±2.1 Vertex AI MaaS Apr 2025 Official leaderboard
8 Gemma 3 27B 15.8% ±1.6 OpenRouter Mar 2025 Official leaderboard

The best score, 60.8%, costs $0.13 a task

Runs on the cost frontier score higher than any run that's cheaper. DeepSeek-V3.2 at Vertex AI MaaS comes within 4.5 points of the best for $0.01 a task, 12 times cheaper. The costs are the leaderboard's own estimates, so compare them only within this benchmark.

Meta Anthropic DeepSeek Cost frontier One model at each effort level
0% 20% 40% 60% 80% $0.001 $0.01 $0.10 $1.00 Cost per task, log scale Claude Sonnet 4, Anthropic API DeepSeek-V3.2, Vertex AI MaaS Llama 4 Scout, Vertex AI MaaS 0% 20% 40% 60% 80% $0.001 $0.01 $0.10 $1.00 Cost per task, log scale Claude Sonnet 4, Anthropic API DeepSeek-V3.2, Vertex AI MaaS

Source: Official GeoNatureAgent Benchmark leaderboard

ModelEffortHarnessScoreCost per taskOutput tokens
Claude Sonnet 4 Anthropic API 60.8% $0.13 557
DeepSeek-V3.2 Vertex AI MaaS 56.3% $0.01 573
GLM-5 Vertex AI MaaS 50.2% $0.04 1,133
Gemini 2.5 Pro Jun 2025 Vertex AI native 48% $0.05 1,587
GPT-4o Mar 2025 OpenRouter 41.6% $0.07 244
gpt-oss-120b Vertex AI MaaS 34.1% $0.09 1,022
Llama 4 Scout Vertex AI MaaS 26.9% $0.003 215
Gemma 3 27B OpenRouter 15.8% $0.06 376

The best score, 60.8%, costs $0.13 a task

Runs on the cost frontier score higher than any run that's cheaper. DeepSeek-V3.2 at Vertex AI MaaS comes within 4.5 points of the best for $0.01 a task, 12 times cheaper. The costs are the leaderboard's own estimates, so compare them only within this benchmark.

Meta Anthropic DeepSeek Cost frontier One model at each effort level
0% 20% 40% 60% 80% $0.001 $0.01 $0.10 $1.00 Cost per task, log scale Claude Sonnet 4, Anthropic API DeepSeek-V3.2, Vertex AI MaaS Llama 4 Scout, Vertex AI MaaS 0% 20% 40% 60% 80% $0.001 $0.01 $0.10 $1.00 Cost per task, log scale Claude Sonnet 4, Anthropic API DeepSeek-V3.2, Vertex AI MaaS

Source: Official GeoNatureAgent Benchmark leaderboard

ModelEffortHarnessScoreCost per taskOutput tokens
Claude Sonnet 4 Anthropic API 60.8% $0.13 557
DeepSeek-V3.2 Vertex AI MaaS 56.3% $0.01 573
GLM-5 Vertex AI MaaS 50.2% $0.04 1,133
Gemini 2.5 Pro Jun 2025 Vertex AI native 48% $0.05 1,587
GPT-4o Mar 2025 OpenRouter 41.6% $0.07 244
gpt-oss-120b Vertex AI MaaS 34.1% $0.09 1,022
Llama 4 Scout Vertex AI MaaS 26.9% $0.003 215
Gemma 3 27B OpenRouter 15.8% $0.06 376

Number one over time

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

Anthropic
Own leaderboard Not tested since Apr 2026

Source: Official GeoNatureAgent 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 Best so far
0% 20% 40% 60% 80% 100% 2025 2026 Near the ceiling (85%) Claude Sonnet 4 GPT-4o Gemma 3 27B 0% 20% 40% 60% 80% 100% 2025 2026 Near the ceiling (85%) Claude Sonnet 4 GPT-4o

Source: Official GeoNatureAgent 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 Tool use

All 265 in Tool use
  • AutomationBench In use Agent completes end-to-end business workflows inside a simulated company exposing dozens of REST APIs. Graded on resulting state. Gemini 4 Neon 56.3%
  • MCP Atlas In use Multi-step workflows over real MCP servers. Scores tool-use competency on realistic tasks. Gemini 4 Argon 90.7%
  • BFCL V4 In use Models call functions and tools, including multi-turn agentic tasks with web search and memory. Checked against expected calls and outcomes. Qwen3.7-Max 75%
  • τ³-bench Banking In use Agents act as banking customer-service reps over many turns with a simulated user, following policy. Pass^1 on the final database state. Qwen 3.8 Max 55.1%
  • Toolathlon-Verified In use 108 long-horizon tasks across 32 apps through about 600 MCP tools. Execution-based checkers inspect final environment state. GLM-5.3-Flash 78.4%
  • DeepSearchQA In use Questions that need multi-step search and synthesis. Final answers graded against references. Kimi K3 95%

Data audit

9 of 10 checks pass, 1 warning. Look first at the Newest score check. The newest came 9 months ago, in February 2026, for GLM-5. Every benchmark

Catalog record 2 pass
  • Pass Task access All tasks are public.
  • Pass Confidence Checked against a primary source.
Scores 1 warning 3 pass 1 note
  • Pass Independent runs Scores from the official leaderboard (9 entries).
  • Pass Models scored 8 models we track, plus 1 other entry on its leaderboard.
  • Warning Newest score The newest came 9 months ago, in February 2026, for GLM-5.
  • Note Frontier models None of the top 10 has a score. The strongest tested is GLM-5 (#89).
  • Pass Headroom The best is 60.8%, under 85%, so it still separates models.
Leaderboard import 3 pass
  • Pass Importer We read the board on huggingface.co: 8 rows for models we track and 1 other entry 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