deepseek-ai/DeepSeek-V4.1-Flash-1017 · Hugging Face
Introduction
DeepSeek-V4.1-Flash-1017 is a minor replace to DeepSeek-V4.1-Flash, delivering substantial enhancements throughout reasoning, coding, agentic duties, and visible benchmarks. Alongside these beneficial properties, the mannequin reduces activity execution time and output size, enabling customers to finish duties quicker and at decrease value and advancing the efficiency–value Pareto frontier throughout a number of agent harnesses.
| Benchmark | DS-V4.1-Flash-1017 | DS-V4.1-Flash | K3 | GLM-5.3 | GPT-5.6 Sol | Opus 5 |
|---|---|---|---|---|---|---|
| HLE w/o instruments | 42.6 (44.0†) | 36.8 (39.1†) | 43.5 | 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/instruments) | 83.0 | 78.9 | 68.1 | — | 79.9 | 84.0 |
| BabyVision (w/instruments) | 92.5 | 89.6 | 85.7 | — | 88.9 | 94.1 |
| ZeroBench-main (w/instruments) | 51.0 | 49.0 | 41.0 | — | 53.0 | 52.0 |
† denotes the text-only subset of HLE. We use the Claude Code harness for Terminal-Bench 4.0 and the Mini-SWE-V2 harness for DeepSWE v1.1.
Prompt Encoding
This launch doesn’t embrace a Jinja-format chat template. The encoding folder incorporates a self-contained Python reference implementation (encoding.py) with check circumstances for multi-turn conversations, instrument calling, pondering mode, numeric reasoning effort, mid-conversation system messages, and interleaved picture content material.
For manufacturing use, we moreover launch deepseek-recipe, a set of Rust libraries with Python bindings that gives the identical immediate format as a maintained, protocol-aware toolkit. It converts Messages, Chat Completions, and Responses API requests into the Conversation format, encodes them into DeepSeek V4 and V4.1 prompts or token IDs, and parses mannequin output again into full or streamed responses — overlaying pondering, instrument calls, pictures, and technology settings. Model inference, instrument execution, and HTTP transport are left to the caller.
Minimal Inference
Please discuss with the inference folder for directions on weight conversion and operating inference regionally.
Recommended sampling parameters:
| Parameter | Value |
|---|---|
temperature |
1.0 |
top_p |
1.0 |
context_window |
1M tokens |
max_tokens |
≥ 256K |
License
This repository and the mannequin weights are licensed underneath the MIT License.
Citation
@misc{deepseekai2026deepseekv41flash,
title={DeepSeek-V4.1-Flash: Pushing the Limits of KV Cache Compression},
writer={DeepSeek-AI},
12 months={2026},
}
Contact
If you’ve gotten any questions, please increase a difficulty or contact us at [email protected].
