DeepSeek V4.1 Flash
MIT-licensed, extremely cheap, and surprisingly strong on code
Overview
DeepSeek V4.1 Flash is a 763B-parameter model with MIT-licensed weights, released in October 2026. Its API price is a small fraction of frontier closed models and it punches above its weight on coding and terminal benchmarks, making it a default pick for high-volume, budget-conscious workloads and self-hosting.
Best for
- High-volume APIs
- Self-hosting
- Coding on a budget
Strengths & limitations
✓ Strengths
- MIT-licensed weights
- Extremely low cost
- Strong terminal/coding scores
△ Limitations
- Weaker on top-end reasoning
- Large self-host footprint
Reported benchmarks
| Benchmark | Score | Note |
|---|---|---|
| Terminal-Bench 2.1 | ~90.6% | vendor-reported |
| GPQA Diamond | ~90.9% | — |
| Artificial Analysis Intelligence Index | ~33–42 | — |
Benchmark figures are reported values from public sources and may use different evaluation harnesses. See sources below.
Frequently asked questions
How much does DeepSeek V4.1 Flash cost?
The API price is $0.15 per 1M input tokens and $0.60 per 1M output tokens (as of 2026-11-02). Prices change — check the official provider page for current rates.
Is DeepSeek V4.1 Flash open source?
DeepSeek V4.1 Flash is open source. Weights are available, subject to the licence terms — review the model card before commercial use.
What is the context window of DeepSeek V4.1 Flash?
DeepSeek V4.1 Flash supports a context window of 1M tokens with a maximum output of 65536 tokens per response.
What is DeepSeek V4.1 Flash best at?
Based on our scoring, DeepSeek V4.1 Flash is strongest at mit-licensed weights, extremely low cost, strong terminal/coding scores. It is a common choice for high-volume apis, self-hosting, coding on a budget.