DeepSeek · 2026-10-17 · major
DeepSeek V4.1 Flash 1017 — an in-place upgrade with stronger agent scores
DeepSeek V4.1 Flash 1017 is a new checkpoint of DeepSeek's open-weight Flash model. It lifts Terminal-Bench 4.0 from 31.2 to 42.3 and replaces the old build on the API, web and app at the same price.
DeepSeek swaps a better-trained checkpoint into deepseek-flash: higher agent and coding scores, shorter outputs, same price.
Key specs
| Terminal bench 4.0 | 42.3 |
|---|---|
| Deep swe v1.1 | 77.0 |
Quick facts
| Maker | DeepSeek |
|---|---|
| Model version | DeepSeek-V4.1-Flash-1017 |
| API model name | deepseek-flash (unchanged) |
| Context window | 1M tokens, 384K max output |
| License | MIT, open weights on Hugging Face |
| Availability | DeepSeek API, web, app, OpenRouter |
Benchmarks
| DeepSeek V4.1 Flash 1017 | 42.3% | |
|---|---|---|
| DeepSeek V4.1 Flash | 31.2% | |
| GLM-5.3 | 37.9% | |
| GPT-5.6 Sol | 39.9% | |
| Opus 5 | 51.8% |
| DeepSeek V4.1 Flash 1017 | 72.1% | |
|---|---|---|
| DeepSeek V4.1 Flash | 64.6% | |
| GLM-5.3 | 56.6% | |
| GPT-5.6 Sol | 54.3% | |
| Opus 5 | 65.8% |
Pricing
| Input · cache miss, off-peak ($0.30 peak) | $0.15 / 1M tokens |
|---|---|
| Input · cache hit, off-peak ($0.006 peak) | $0.003 / 1M tokens |
| Output · off-peak ($1.20 peak) | $0.60 / 1M tokens |
What is it?
DeepSeek-V4.1-Flash-1017 replaced the September build of DeepSeek V4.1 Flash on October 17 across the API, the web chat and the app. The model card calls it a minor update with large gains on reasoning, coding, agent and vision benchmarks. The 1017 weights are on Hugging Face and ModelScope under the same MIT license.
How does it work?
DeepSeek says the checkpoint was trained with reinforcement learning with self-feedback, aimed at building whole projects end to end; its demos include a 3D Forbidden City and a Portal 2 recreation in Godot. The architecture and 1M-token context carry over from V4.1 Flash. On DeepSeek's own table every benchmark improved, the biggest jumps being Terminal-Bench 4.0 (+11.1) and ProgramBench (+8.8).
Why does it matter?
Anyone already calling deepseek-flash gets the upgrade with no code change and no price change, and DeepSeek says tasks now finish faster with shorter outputs, which cuts cost per task. On the published table the cheap open model now beats GPT-5.6 Sol on 9 of 11 benchmarks, though Opus 5 still leads most rows. The same release opened DeepSeek's in-house Search and Fetch APIs to everyone.
Who is it for?
developers on the DeepSeek API and teams self-hosting open-weight models
Frequently asked questions
- Do I need to change my code to use DeepSeek V4.1 Flash 1017?
- No code change is needed for DeepSeek V4.1 Flash 1017. DeepSeek's changelog says the deepseek-flash alias now serves the 1017 checkpoint, and pricing and calling method are unchanged. The legacy names deepseek-v4-flash and deepseek-v4-flash-vision-exp also route to the 1017 model at Flash prices. DeepSeek recommends DeepSeek Harness v0.2.1-rc.1 or later for agent work.
- How much does DeepSeek V4.1 Flash 1017 cost?
- DeepSeek V4.1 Flash 1017 keeps V4.1 Flash pricing: $0.15 per million input tokens on a cache miss and $0.60 per million output tokens off-peak, doubling to $0.30 and $1.20 in peak hours. Cache hits cost $0.003 off-peak. Peak hours are 01:00–04:00 and 06:00–10:00 UTC on weekdays, excluding Chinese public holidays.
- How does DeepSeek V4.1 Flash 1017 compare with Opus 5 and GPT-5.6 Sol?
- On DeepSeek's published model-card table, DeepSeek V4.1 Flash 1017 beats GPT-5.6 Sol on 9 of 11 benchmarks, trailing only on HLE without tools (42.6 vs 44.5) and ZeroBench. Opus 5 still leads most rows, including Terminal-Bench 4.0 (51.8 vs 42.3), but the 1017 checkpoint is ahead of it on DeepSWE v1.1, Automation-Bench, Agents' Last Exam and Apex Bench.
- What are DeepSeek's new Search and Fetch APIs?
- DeepSeek opened its in-house Search and Fetch APIs to everyone in the same October 17 announcement as DeepSeek V4.1 Flash 1017. Each endpoint costs $0.75 per 1,000 calls. Calls made through server-side tools in DeepSeek Harness, Claude Code and Codex are free until billing starts on October 31, 2026 at 16:00 UTC.
Try it
model="deepseek-flash" on https://api.deepseek.com