Mistral Large 4 (high) vs Claude Haiku 4.5 (Reasoning)
Results updated 3 November 2026 · In Terminus-2 and Claude Code, so part of any gap can come from the harness · How we benchmark
Verdict
Mistral Large 4 (high) edges Claude Haiku 4.5 (Reasoning) by 1 point; Claude Haiku 4.5 (Reasoning) is cheaper and faster.
- Cost per task
- $0.10 vs $0.092
- Time per task
- 12.0 min vs 1.6 min
- Tokens per attempt
- 226k vs 313k
- Image input
- Yes vs Yes
Mistral Large 4 (high), Terminus-2 scores higher on GeoBenchX and GS-QA2; Claude Haiku 4.5 (Reasoning), Claude Code scores higher on Earth-Bench V2. Claude Haiku 4.5 (Reasoning), Claude Code is cheaper and faster: $0.092 against $0.10 per task, and 1.6 min against 12.0 min.
Where they differ most
By kind of work, share of tasks passed, 0 to 100 · Higher is better
| Kind of work | Mistral Large 4 (high) | Claude Haiku 4.5 (Reasoning) |
|---|---|---|
| Recognising infeasible requests | 23 | 53 |
| Making maps | 82 | 56 |
| Spatial statistics | 46 | 23 |
| GIS analysis | 82 | 67 |
Where the tasks are set: each kind of work links to its tasks and scoresCharts for each: Mistral Large 4 (high), Terminus-2 · Claude Haiku 4.5 (Reasoning), Claude Code.
Side by side
| Metric | Mistral Large 4 (high), Terminus-2 | Claude Haiku 4.5 (Reasoning), Claude Code |
|---|---|---|
| Geospatial Agent Index: Text only | 59 | 58 |
| Geospatial Agent Index: Text + vision | No data | 48 |
| Earth-Bench V2 score | 46 | 51 |
| GeoBenchX score | 55 | 53 |
| GS-QA2 score | 77 | 69 |
| Vision agents score | No data | 19 |
| GeoAgentBench score (Guided workflows; not in the Agent Index) | 80 | 78 |
| GeoAnalystBench score (Guided workflows; not in the Agent Index) | 94 | 74 |
| OGD4All score (standalone; in no index) | 96 | 95 |
| GeoNatureAgent score (standalone; in no index) | 100 | 93 |
| MultiGlobeQA score (standalone; in no index) | 62 | 70 |
| GeoAgentEval score (standalone; in no index) | 60 | 88 |
| Remote sensing and raster analysis score | 67 | 65 |
| GIS analysis score | 82 | 67 |
| Climate and time series score | 40 | 49 |
| Spatial statistics score | 46 | 23 |
| Making maps score | 82 | 56 |
| Recognising infeasible requests score | 23 | 53 |
| Cost per task | $0.10 | $0.092 |
| Time per task (at 100 output tokens/s) | 12.0 min | 1.6 min |
| Tokens per attempt | 226k | 313k |
| Turns per attempt | 13.8 | 10.6 |
| Input price per 1M tokens | $0.70 | $1.00 |
| Output price per 1M tokens | $2.10 | $5.00 |
| Context window | 262,144 | 200,000 |
| Image input | Yes | Yes |
| Developer | Mistral AI | Anthropic |
Frequently asked questions
Which scores higher on the Geospatial Agent Index, Mistral Large 4 (high), Terminus-2 or Claude Haiku 4.5 (Reasoning), Claude Code?
Mistral Large 4 (high), Terminus-2, 59 to 58.
Which is cheaper per task, Mistral Large 4 (high), Terminus-2 or Claude Haiku 4.5 (Reasoning), Claude Code?
Claude Haiku 4.5 (Reasoning), Claude Code, $0.092 to $0.10.
Which writes fewer output tokens per task, Mistral Large 4 (high), Terminus-2 or Claude Haiku 4.5 (Reasoning), Claude Code?
Claude Haiku 4.5 (Reasoning), Claude Code, 7k to 14k.
Which has the larger context window, Mistral Large 4 (high), Terminus-2 or Claude Haiku 4.5 (Reasoning), Claude Code?
Mistral Large 4 (high), Terminus-2, 262,144 tokens to 200,000 tokens.