In short
How they differ.
Worked out from the figures below. Prices and specs are the makers' own, read Oct 20, 2026.
- GPT-6 Astra reads more at once: a context window of 1M tokens against 256K.
- Gemma 4 31B can be downloaded and run on your own machine; GPT-6 Astra can't.
- No independent board we follow ranks both models yet.
Side by side
The specs.
| Gemma 4 31B | GPT-6 Astra | |
|---|---|---|
| Made by | Google · US | OpenAI · US |
| Released | Apr 2, 2026 | Sep 3, 2026 |
| Price per 1M tokens | Free to download; hosts set prices | $10 in / $50 out |
| Context window | 256K | 1M |
| Takes in | text, image, video | text, image |
| Download and run it yourself | Yes (Apache 2.0) | No |
| Size | 30.7B (dense) parameters | Not published |
| Maker says | “Today, we are introducing Gemma 4 — our most intelligent open models to date.” | “Our most capable model for the most demanding work.” |
| Sources | Prices, Oct 20, 2026 · Description | Prices, Oct 20, 2026 · Description |
Independent tests
Where they place.
Position on each board we follow (1 is best). Epoch AI's and Vectara's scores are shown; the other boards are linked for their own figures.
| Board | Gemma 4 31B | GPT-6 Astra |
|---|---|---|
| Epoch Capabilities IndexEpoch AI, Oct 20, 2026 | Not listed | #2166.41 |
| LMArena coding chatsLMArena, Oct 15, 2026 | Not listed | #9 |
| LMArena WebDevLMArena, Oct 19, 2026 | Not listed | #2 |
| LMArena documentsLMArena, Sep 13, 2026 | Not listed | #16 |
| LMArena Agent ArenaLMArena, Oct 19, 2026 | Not listed | #2 |
| Humanity's Last ExamScale AI | Not listed | #1 |
| SWE-Bench ProScale AI, Sep 22, 2026 | Not listed | #5 |
| Remote Labor IndexScale AI | Not listed | #2 |
| Vals IndexVals AI, Oct 19, 2026 | Not listed | #5 |
| Terminal-Bench 4.0Vals AI, Oct 19, 2026 | Not listed | #3 |
| Vibe Code BenchVals AI, Oct 18, 2026 | Not listed | #7 |
| ProofBenchVals AI, Oct 12, 2026 | Not listed | #6 |
| CUA-bench (games)Vals AI, Oct 13, 2026 | Not listed | #1 |
| Vectara hallucination testVectara, Oct 16, 2026 | Not listed | #458.7% |
More comparisons