GLM-5.3 (max) vs DeepSeek V4.1 Flash (max)
Results updated 31 October 2026 · In Terminus-2 and Claude Code, so part of any gap can come from the harness · How we benchmark
Verdict
GLM-5.3 (max) edges DeepSeek V4.1 Flash (max) by 2 points, and is faster; DeepSeek V4.1 Flash (max) is cheaper.
- Cost per task
- $0.38 vs $0.20
- Time per task
- 4.2 min vs 10.9 min
- Tokens per attempt
- 278k vs 1.58M
- Image input
- No vs Yes
GLM-5.3 (max), Terminus-2 scores higher on Earth-Bench V2 and GeoBenchX; DeepSeek V4.1 Flash (max), Claude Code scores higher on GS-QA2. DeepSeek V4.1 Flash (max), Claude Code is cheaper ($0.20 against $0.38 per task), but GLM-5.3 (max), Terminus-2 is faster (4.2 min against 10.9 min).
Where they differ most
By kind of work, share of tasks passed, 0 to 100 · Higher is better
| Kind of work | GLM-5.3 (max) | DeepSeek V4.1 Flash (max) |
|---|---|---|
| Spatial statistics | 54 | 31 |
| Recognising infeasible requests | 55 | 40 |
| GIS analysis | 80 | 92 |
| Climate and time series | 71 | 64 |
Where the tasks are set: each kind of work links to its tasks and scoresCharts for each: GLM-5.3 (max), Terminus-2 · DeepSeek V4.1 Flash (max), Claude Code.
Side by side
| Metric | GLM-5.3 (max), Terminus-2 | DeepSeek V4.1 Flash (max), Claude Code |
|---|---|---|
| Geospatial Agent Index: Text only | 70 | 68 |
| Geospatial Agent Index: Text + vision | No image input | No data |
| Earth-Bench V2 score | 64 | 58 |
| GeoBenchX score | 65 | 55 |
| GS-QA2 score | 83 | 89 |
| Vision agents score | No image input | No data |
| GeoAgentBench score (Guided workflows; not in the Agent Index) | 94 | 96 |
| GeoAnalystBench score (Guided workflows; not in the Agent Index) | 84 | 100 |
| OGD4All score (standalone; in no index) | No data | 79 |
| GeoNatureAgent score (standalone; in no index) | No data | 100 |
| MultiGlobeQA score (standalone; in no index) | No data | 64 |
| GeoAgentEval score (standalone; in no index) | No data | 84 |
| Remote sensing and raster analysis score | 76 | 75 |
| GIS analysis score | 80 | 92 |
| Climate and time series score | 71 | 64 |
| Spatial statistics score | 54 | 31 |
| Making maps score | 75 | 72 |
| Recognising infeasible requests score | 55 | 40 |
| Cost per task | $0.38 | $0.20 |
| Time per task (at 100 output tokens/s) | 4.2 min | 10.9 min |
| Tokens per attempt | 278k | 1.58M |
| Turns per attempt | 10.4 | 26.0 |
| Input price per 1M tokens | $1.40 | $0.30 |
| Output price per 1M tokens | $4.40 | $1.20 |
| Context window | 1,048,576 | 1,048,576 |
| Image input | No | Yes |
| Developer | Z.ai | DeepSeek |
Frequently asked questions
Which scores higher on the Geospatial Agent Index, GLM-5.3 (max), Terminus-2 or DeepSeek V4.1 Flash (max), Claude Code?
GLM-5.3 (max), Terminus-2, 70 to 68.
Which is cheaper per task, GLM-5.3 (max), Terminus-2 or DeepSeek V4.1 Flash (max), Claude Code?
DeepSeek V4.1 Flash (max), Claude Code, $0.20 to $0.38.
Which writes fewer output tokens per task, GLM-5.3 (max), Terminus-2 or DeepSeek V4.1 Flash (max), Claude Code?
GLM-5.3 (max), Terminus-2, 19k to 48k.
Which has the larger context window, GLM-5.3 (max), Terminus-2 or DeepSeek V4.1 Flash (max), Claude Code?
They are level, at 1,048,576 tokens.