In short
How they differ.
Worked out from the figures below. Prices and specs are the makers' own, read Oct 20, 2026.
- DeepSeek-V4-Pro-0813 is cheaper per output token: $3.96 against $6 per million tokens, so Grok 4.7 costs 1.5 times as much.
- DeepSeek-V4-Pro-0813 reads more at once: a context window of 1M tokens against 500K.
- DeepSeek-V4-Pro-0813 can be downloaded and run on your own machine; Grok 4.7 can't.
- On the 2 independent boards that rank both, Grok 4.7 places higher on 2 and DeepSeek-V4-Pro-0813 on 0.
Side by side
The specs.
| DeepSeek-V4-Pro-0813 | Grok 4.7 | |
|---|---|---|
| Made by | DeepSeek · China | SpaceXSI (formerly xAI) · US |
| Released | Aug 13, 2026 | Sep 21, 2026 |
| Price per 1M tokens | $1.32 in / $3.96 out (peak-hours price) | $2 in / $6 out |
| Context window | 1M | 500K |
| Takes in | text | text, image |
| Download and run it yourself | Yes (MIT) | No |
| Size | 1.6T parameters, 49B active | Not published |
| Maker says | — | “Grok 4.7 is our most capable model for coding and knowledge work.” |
| Sources | Prices, Oct 20, 2026 | 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 | DeepSeek-V4-Pro-0813 | Grok 4.7 |
|---|---|---|
| LMArena WebDevLMArena, Oct 19, 2026 | Not listed | #18 |
| LMArena Agent ArenaLMArena, Oct 19, 2026 | Not listed | #17 |
| Humanity's Last ExamScale AI | #16 | #13 |
| Vals IndexVals AI, Oct 19, 2026 | #17 | #11 |
| Terminal-Bench 4.0Vals AI, Oct 19, 2026 | Not listed | #20 |
| Vibe Code BenchVals AI, Oct 18, 2026 | Not listed | #12 |
| GPQA Diamond (run by Vals AI, archived)Vals AI, Sep 1, 2026 | #17 | Not listed |
| Vectara hallucination testVectara, Oct 16, 2026 | #448.6% | Not listed |
More comparisons