DeepSeek
DeepSeek V4
DeepSeek V4.1 Flash-1017 updates reasoning, coding, agent, and vision performance on October 17, 2026. The deepseek-flash API model ID and pricing stay unchanged; V4 Pro-0813 remains separately available.
Flash release
1017
Official October 17, 2026 update to V4.1 Flash.
Context
1M
V4.1 Flash supports text and image inputs with a one-million-token context window.
API model ID
deepseek-flash
The model ID and API pricing remain unchanged for this release.
Model Overview
DeepSeek V4.1 Flash-1017 and V4 Pro-0813; benchmark results are labeled by version.
Params
Pro 1.6T · V4.1 Flash 552B
Context
1M
Released
Flash: Oct 17 · Pro: Aug 13, 2026
DeepSeek V4.1 Flash-1017: October 17 release benchmarks
DeepSeek reports gains across reasoning, coding, agent tasks, and vision. This release table compares Flash-1017 with the September Flash release and four other models; the scores are publisher-reported, not an independent evaluation. The API model ID remains deepseek-flash and pricing is unchanged.
| Benchmark | DeepSeek-V4.1-Flash-1017 | DeepSeek-V4.1-Flash (2026-09-10) | Kimi K3 | GLM-5.3 | GPT-5.6 Sol | Claude Opus 5.0 |
|---|---|---|---|---|---|---|
| HLE (w/o tools) | 42.6 (44.0†) | 36.8 (39.1†) | 43.5 | Not reported (42.0†) | 44.5 | 56.3 |
| Terminal-Bench 4.0 | 42.3 | 31.2 | 12.6 | 37.9 | 39.9 | 51.8 |
| DeepSWE v1.1 | 77.0 | 74.2 | 67.5 | 66.9 | 73.0 | 74.0 |
| ProgramBench | 29.1 | 20.3 | 17.5 | 19.0 | 23.0 | 37.0 |
| NL2Repo-Bench | 68.2 | 65.4 | 58.0 | 58.0 | 56.8 | 75.3 |
| Automation-Bench v1.0.6 | 60.5 | 54.8 | 46.7 | 48.8 | 45.8 | 50.3 |
| Agents’ Last Exam | 35.4 | 31.8 | 27.6 | 28.5 | 26.7 | 28.6 |
| Apex Bench v1.1 | 72.1 | 64.6 | 47.8 | 56.6 | 54.3 | 65.8 |
| Chartography (w/tools) | 83.0 | 78.9 | 68.1 | Not reported | 79.9 | 84.0 |
| BabyVision (w/tools) | 92.5 | 89.6 | 85.7 | Not reported | 88.9 | 94.1 |
| ZeroBench-main (w/tools, Pass@5) | 51.0 | 49.0 | 41.0 | Not reported | 53.0 | 52.0 |
† HLE text-only subset, not the full test. GLM-5.3 has only a text-only HLE result in this source. Terminal-Bench 4.0 uses Claude Code; DeepSWE v1.1 uses Mini-SWE-V2. Chartography, BabyVision, and ZeroBench-main use tools; ZeroBench-main is Pass@5. Automation-Bench v1.0.6 and Apex Bench v1.1 retain their published versions. No new Pro-0813 or Terminal-Bench 2.1 result is supplied in this release table.
Source: DeepSeek Flash-1017 model card · October 17, 2026 · Official model card · Official changelog
Model overview
Written in a launch-style profile format
DeepSeek's October 17 model card compares Flash-1017 with the September V4.1 Flash release and four other models across 11 benchmarks. These are publisher-reported results, with the full table and evaluation notes shown above.
The release also reports shorter task runtimes and less output. Actual cost and latency depend on the task, effort, and harness; the announcement does not provide a universal speedup. DeepSeek recommends Harness 0.2.1-rc.1 or later for this update.
Model highlights
Key strengths and deployment profile
Reasoning and coding
The update includes HLE without tools, Terminal-Bench 4.0, DeepSWE v1.1, ProgramBench, and NL2Repo-Bench. Keep each benchmark version and harness attached to its scores.
Multimodal agents
V4.1 Flash supports text and image inputs. Chartography, BabyVision, and ZeroBench-main are evaluated with tools; ZeroBench-main uses Pass@5.
Developer experience
Existing deepseek-flash requests reach Flash-1017 without a model-ID or token-price change. Keep the current API configuration and update DeepSeek Harness to 0.2.1-rc.1 or later when using that client.
Separate model versions
The September Flash baseline remains a historical comparison. V4 Pro-0813 stays separately available, but the October 17 model card supplies no new Pro scores. Choose using results for your actual task and model version.
See what it can do
See what DeepSeek V4 can do
Built for developers
Give developers more control
Best For
Who should start with this model
When To Choose It
When it belongs on your shortlist
Current Coding Plan Inventory