DeepSeek V4.1 Flash 1017 Goes Free
Experiential Labs is temporarily offering DeepSeek-V4.1-Flash-1017 free through its gateway, with a 1.05M-token context window and up to 393K output tokens. DeepSeek says the update improves coding and agentic performance while completing tasks faster with shorter outputs.
This is a compelling short-term testbed for long-running coding agents, but developers should treat the free lane as a promotion rather than dependable production infrastructure.
- –The massive context and output limits suit repository-scale analysis, multi-step coding, and project generation.
- –The model card reports gains over the previous Flash version, including Terminal-Bench improvement from 31.2 to 42.3 and DeepSWE improvement from 74.2 to 77.0. [Hugging Face](https://huggingface.co/deepseek-ai/DeepSeek-V4.1-Flash-1017)
- –Experiential Labs’ gateway makes evaluation easy through an OpenAI-compatible API, lowering the friction for agent and IDE experiments.
- –DeepSeek has also published the model on Hugging Face and ModelScope, giving teams options beyond the temporary hosted giveaway. [DeepSeek](https://api-docs.deepseek.com/news/news261017/)
DISCOVERED
1h ago
2026-10-17
PUBLISHED
2h ago
2026-10-17
RELEVANCE
AUTHOR
0x0SojalSec
