DeepSeek V4.1 Flash, MiniMax M3.1 Flash, Gemini 4 Argon
Two low-cost “Flash” models against Google's frontier model: price per million tokens, DeepSWE v1.1 score, context, open weights and release date.
| Compared tools | Vendor | Input price ($/M tokens) | Output price ($/M tokens) | Coding score | Context (tokens) | Open weights | Release |
|---|---|---|---|---|---|---|---|
| DeepSeek V4.1-Flash-1017 | DeepSeek | $0.3 | $1.2 | 77% (DeepSWE v1.1) | 1,000,000 | Yes | |
| MiniMax M3.1 Flash | MiniMax | $0.15 | $0.6 | 73.7% (DeepSWE v1.1) | 1,000,000 | No | |
| Gemini 4 Argon | $2 | $10 | 77.9% (DeepSWE v1.1) | 1,000,000 | No |
Standard list prices in US dollars per million tokens (DeepSeek: peak hours, half price off-peak; Gemini 4 Argon: launch price until November 30, 2026, then $4/$20). DeepSWE v1.1 scores are announced by each vendor under its own test conditions: they are not strictly comparable.
Sources
Verified
- Hugging Face: model card (MIT licence)
- DeepSeek: October 17, 2026 announcement
- DeepSeek: API pricing
- MiniMax: M3.1 Flash announcement
- MiniMax: model documentation (context)
- Google: Gemini 4 Argon announcement
- Google: Gemini API pricing
- Google: Gemini API models (context)
- Google: Gemini 4 Argon introduction (DeepSWE v1.1, Sep 30, 2026)