DeepSeek-V4.1-Flash deepseek-ai

Status

1. Weights
2. Providers
3. Quants
4. Pricing
5. Articles

Weights → providers → quants → pricing → articles. A step turns on when the engine observes it; it never turns on from an announcement.

Timeline

1. 2026-09-10 · Weights published on Hugging Face source
2. 2026-09-10 · Repo commit: Update inference/README.md source
3. 2026-09-10 · Repo commit: Update README.md & encoding/README.md source
4. 2026-09-10 · Repo commit: Update README.md & encoding/README.md source
5. 2026-09-10 · Repo commit: fix(encoding): preserve tool namespaces in prompts and completions source
6. 2026-09-10 · Article: DeepSeek v4.1 Flash source
7. 2026-09-16 · Article: DeepSeek v4.1 Flash Is Now Our Best Hacking Model source
8. 2026-09-17 · Article: DeepSeek-v4.1 Flash: Pushing the Limits of KV Cache Compression source
9. 2026-10-01 · Repo commit: Adding chat template (#68) source
10. 2026-10-12 · Repo commit: Restore each indexer's K cache on incomplete compression steps (#12) source
11. 2026-10-12 · Repo commit: update source
12. 2026-10-12 · Repo commit: Delete files inference/__pycache__/ with huggingface_hub source
13. 2026-10-12 · Repo commit: gitignore source
14. 2026-10-16 · Repo commit: fix(encoding): preserve tool image order and current-turn thinking source
15. 2026-10-25 · 34 hosted providers observed source
16. 2026-10-25 · 137 community quantizations on the Hub source

Watch via RSS