MiniMax M3 (Reasoning) vs GPT-5.6 Luna (low)
Results updated 8 November 2026 · In MiniMax Code and Codex, so part of any gap can come from the harness · How we benchmark
Verdict
MiniMax M3 (Reasoning) beats GPT-5.6 Luna (low) by 7 points; GPT-5.6 Luna (low) is cheaper and faster.
- Cost per task
- $0.15 vs $0.0058
- Time per task
- 5.0 min (estimated) vs 0.5 min
- Tokens per attempt
- 1.69M vs 84k
- Image input
- Yes vs Yes
MiniMax M3 (Reasoning), MiniMax Code scores higher on Earth-Bench V2 and GS-QA2; GPT-5.6 Luna (low), Codex scores higher on GeoBenchX. GPT-5.6 Luna (low), Codex is cheaper and faster: $0.0058 against $0.15 per task, and 0.5 min against 5.0 min.
Where they differ most
By kind of work, share of tasks passed, 0 to 100 · Higher is better
| Kind of work | MiniMax M3 (Reasoning) | GPT-5.6 Luna (low) |
|---|---|---|
| Climate and time series | 72 | 40 |
| Change detection | 64 | 52 |
| GIS analysis | 81 | 72 |
| Making maps | 77 | 81 |
Where the tasks are set: each kind of work links to its tasks and scoresCharts for each: MiniMax M3 (Reasoning), MiniMax Code · GPT-5.6 Luna (low), Codex.
Side by side
| Metric | MiniMax M3 (Reasoning), MiniMax Code | GPT-5.6 Luna (low), Codex |
|---|---|---|
| Geospatial Agent Index: Text only | 68 | 61 |
| Geospatial Agent Index: Text + vision | No data | 62 |
| Earth-Bench V2 score | 68 | 44 |
| GeoBenchX score | 59 | 62 |
| GS-QA2 score | 78 | 76 |
| Vision agents score | No data | 64 |
| GeoAgentBench score (Guided workflows; not in the Agent Index) | 94 | 86 |
| GeoAnalystBench score (Guided workflows; not in the Agent Index) | 79 | 95 |
| OGD4All score (standalone; in no index) | 94 | No data |
| GeoNatureAgent score (standalone; in no index) | 93 | No data |
| MultiGlobeQA score (standalone; in no index) | 73 | No data |
| GeoAgentEval score (standalone; in no index) | 92 | No data |
| ImpactMesh score (standalone; in no index) | No data | No data |
| Remote sensing and raster analysis score | 74 | 72 |
| Change detection score | 64 | 52 |
| GIS analysis score | 81 | 72 |
| Climate and time series score | 72 | 40 |
| Making maps score | 77 | 81 |
| Cost per task | $0.15 | $0.0058 |
| Time per task (at 100 output tokens/s) | 5.0 min (estimated) | 0.5 min |
| Tokens per attempt | 1.69M | 84k |
| Turns per attempt | No data | 7.0 |
| Input price per 1M tokens | $0.30 | $0.20 |
| Output price per 1M tokens | $1.20 | $1.20 |
| Context window | 1,048,576 | 1,050,000 |
| Image input | Yes | Yes |
| Developer | MiniMax | OpenAI |
Frequently asked questions
Which scores higher on the Geospatial Agent Index, MiniMax M3 (Reasoning), MiniMax Code or GPT-5.6 Luna (low), Codex?
MiniMax M3 (Reasoning), MiniMax Code, 68 to 61.
Which is cheaper per task, MiniMax M3 (Reasoning), MiniMax Code or GPT-5.6 Luna (low), Codex?
GPT-5.6 Luna (low), Codex, $0.0058 to $0.15.
Which writes fewer output tokens per task, MiniMax M3 (Reasoning), MiniMax Code or GPT-5.6 Luna (low), Codex?
GPT-5.6 Luna (low), Codex, 2k to 30k.
Which has the larger context window, MiniMax M3 (Reasoning), MiniMax Code or GPT-5.6 Luna (low), Codex?
GPT-5.6 Luna (low), Codex, 1,050,000 tokens to 1,048,576 tokens.