Alex Getmanearly AI signals
EN

DeepSeek V4.1 Flash

DeepSeek

Released Updated
Context window
1M
1,048,576 tokens
Input / Output
image, text → text

Timeline

  1. · Weights version · evidence link
  1. · Evidence links · GitHub search · DeepSeek · commits · vhaldemar/opencode-plugin-yandex-ai-studio · @vhaldemar
  2. · Evidence links · GitHub search · DeepSeek · commits · roasaetn2-crypto/solana-desk- · @unknown
    Excerpt
    Default deepseek-v4.1-flash with four fallbacks; RPM 6, budget 200.
  3. · Evidence links · GitHub search · DeepSeek · commits · ricardoblackskye/agent-eve · @hermes-backup
  4. · Evidence links · GitHub search · DeepSeek · commits · ricardoblackskye/agent-eve · @hermes-backup
  5. · Evidence links · GitHub search · DeepSeek · commits · nico359/camera · @nico359
  6. · Evidence links · GitHub search · DeepSeek · commits · nico359/camera · @nico359
  7. · Evidence links · GitHub search · DeepSeek · commits · nice-bills/bonded · @nice-bills
  8. · Evidence links · GitHub search · DeepSeek · commits · mhqdz/Xaihi-AI · @mhqdz
  9. · Evidence links · GitHub search · DeepSeek · commits · juliolv/portfolio · @juliolv
  10. · Evidence links · GitHub search · DeepSeek · commits · juliolv/portfolio · @juliolv
  11. · Evidence links · GitHub search · DeepSeek · commits · juliolv/portfolio · @juliolv
    Excerpt
    paid model; the cheap tier (deepseek-v4.1-flash, glm-5.3-flash, qwen3.8-flash,
  12. · Evidence links · GitHub search · DeepSeek · commits · guidemeextension-ai/rayucode-e-ticket · @sassdahRAK
  13. · Evidence links · GitHub search · DeepSeek · commits · earino/dataset-factory · @unknown
    Excerpt
    ollama-cloud, model deepseek-v4.1-flash (previously nous). The host
  14. · Evidence links · GitHub search · DeepSeek · commits · doggy8088/ask-page-extension · @doggy8088
  15. · Evidence links · GitHub search · DeepSeek · commits · bloomdevelop/bloomdevelop.github.io · @bloomdevelop
  16. · Evidence links · GitHub search · DeepSeek · commits · bloomdevelop/bloomdevelop.github.io · @bloomdevelop
  17. · Evidence links · GitHub search · DeepSeek · commits · athal7/dotfiles · @athal7
  18. · Evidence links · GitHub search · DeepSeek · commits · Moosemorals/uk.osric.asp · @Moosemorals
    Excerpt
    500,000 tokens, the output limit to the 393,216 DeepSeek-V4.1-Flash reports, and a tool result to 131,072 characters. The step and tool-call histograms reach past their
  19. · Evidence links · GitHub search · DeepSeek · pulls · yonatan895/qdrant-pdf-rag · @yonatan895
  20. · Evidence links · GitHub search · DeepSeek · pulls · y5ez3g/facadecode · @y5ez3g
  21. · Evidence links · GitHub search · DeepSeek · pulls · voxix-dev/justpass-ai · @voxix-dev
  22. · Evidence links · GitHub search · DeepSeek · pulls · vercel-labs/fx · @fazxes
    Excerpt
    - Switch that run to `deepseek/deepseek-v4.1-flash` once the main agent is idle when the team has an active Token Pass, without saving it as the
  23. · Evidence links · GitHub search · DeepSeek · pulls · veniceai/api-docs · @sabrinaaquino
  24. · Evidence links · GitHub search · DeepSeek · pulls · uwuclxdy/ai-pricelog · @cldbot
  25. · Evidence links · GitHub search · DeepSeek · pulls · sgl-project/sglang · @rainj-me
    Excerpt
    against the Python frontend side by side on DeepSeek-V4.1-Flash (both frontends TP2 EP2 on one GB300 tray); the remaining differences are pre-existing main-vs-branch gaps or
  26. · Evidence links · GitHub search · DeepSeek · pulls · sgl-project/sglang · @rainj-me
    Excerpt
    against the Python frontend side by side on DeepSeek-V4.1-Flash (both frontends TP2 EP2 on one GB300 tray); the remaining differences are pre-existing main-vs-branch gaps or
  27. · Evidence links · GitHub search · DeepSeek · pulls · sgl-project/sglang · @rainj-me
    Excerpt
    against the Python frontend side by side on DeepSeek-V4.1-Flash (both frontends TP2 EP2 on one GB300 tray); the remaining differences are pre-existing main-vs-branch gaps or
  28. · Evidence links · GitHub search · DeepSeek · pulls · sgl-project/sglang · @rainj-me
    Excerpt
    against the Python frontend side by side on DeepSeek-V4.1-Flash (both frontends TP2 EP2 on one GB300 tray); the remaining differences are pre-existing main-vs-branch gaps or
  29. · Evidence links · GitHub search · DeepSeek · pulls · sgl-project/sglang · @rainj-me
    Excerpt
    against the Python frontend side by side on DeepSeek-V4.1-Flash (both frontends TP2 EP2 on one GB300 tray); the remaining differences are pre-existing main-vs-branch gaps or
  30. · Evidence links · GitHub search · DeepSeek · pulls · sgl-project/sglang · @rainj-me
    Excerpt
    against the Python frontend side by side on DeepSeek-V4.1-Flash (both frontends TP2 EP2 on one GB300 tray); the remaining differences are pre-existing main-vs-branch gaps or
  31. · Evidence links · GitHub search · DeepSeek · pulls · sgl-project/sglang · @rainj-me
    Excerpt
    against the Python frontend side by side on DeepSeek-V4.1-Flash (both frontends TP2 EP2 on one GB300 tray); the remaining differences are pre-existing main-vs-branch gaps or
  32. · Evidence links · GitHub search · DeepSeek · pulls · sgl-project/sglang · @rainj-me
    Excerpt
    against the Python frontend side by side on DeepSeek-V4.1-Flash (both frontends TP2 EP2 on one GB300 tray); the remaining differences are pre-existing main-vs-branch gaps or
  33. · Evidence links · GitHub search · DeepSeek · pulls · sgl-project/sglang · @rainj-me
    Excerpt
    against the Python frontend side by side on DeepSeek-V4.1-Flash (both frontends TP2 EP2 on one GB300 tray); the remaining differences are pre-existing main-vs-branch gaps or
  34. · Evidence links · GitHub search · DeepSeek · pulls · sgl-project/sglang · @rainj-me
    Excerpt
    against the Python frontend side by side on DeepSeek-V4.1-Flash (both frontends TP2 EP2 on one GB300 tray); the remaining differences are pre-existing main-vs-branch gaps or
  35. · Evidence links · GitHub search · DeepSeek · pulls · sgl-project/sglang · @rainj-me
    Excerpt
    against the Python frontend side by side on DeepSeek-V4.1-Flash (both frontends TP2 EP2 on one GB300 tray); the remaining differences are pre-existing main-vs-branch gaps or
  36. · Evidence links · GitHub search · DeepSeek · pulls · sgl-project/sglang · @rainj-me
    Excerpt
    against the Python frontend side by side on DeepSeek-V4.1-Flash (both frontends TP2 EP2 on one GB300 tray); the remaining differences are pre-existing main-vs-branch gaps or
  37. · Evidence links · GitHub search · DeepSeek · pulls · sgl-project/sglang · @rainj-me
    Excerpt
    against the Python frontend side by side on DeepSeek-V4.1-Flash (both frontends TP2 EP2 on one GB300 tray); the remaining differences are pre-existing main-vs-branch gaps or
  38. · Evidence links · GitHub search · DeepSeek · pulls · sgl-project/sglang · @rainj-me
    Excerpt
    against the Python frontend side by side on DeepSeek-V4.1-Flash (both frontends TP2 EP2 on one GB300 tray); the remaining differences are pre-existing main-vs-branch gaps or
  39. · Evidence links · GitHub search · DeepSeek · pulls · sgl-project/sglang · @rainj-me
    Excerpt
    against the Python frontend side by side on DeepSeek-V4.1-Flash (both frontends TP2 EP2 on one GB300 tray); the remaining differences are pre-existing main-vs-branch gaps or
  40. · Evidence links · GitHub search · DeepSeek · pulls · sgl-project/sglang · @amd-pedghazi
  41. · Evidence links · GitHub search · DeepSeek · pulls · ricardoblackskye/agent-eve · @ricardoblackskye
    Excerpt
    1_048_576), so a single turn reserved **$1.2156** on `deepseek/deepseek-v4.1-flash`. `admit()` refuses when `spent + reserved + estimate > cap`, so once a period's headroom fell
  42. · Evidence links · GitHub search · DeepSeek · pulls · professorpalmer/marionette · @professorpalmer
  43. · Evidence links · GitHub search · DeepSeek · pulls · phase-rs/phase · @howeasy
    Excerpt
    claude-opus-5-5; claude-sonnet-5-5; gpt-6.1-sol (via OMP; reported as openai-codex/gpt-6.1-sol); opencode-go/deepseek-v4.1-flash (via OMP; canonical id not exposed)
  44. · Evidence links · GitHub search · DeepSeek · pulls · phase-rs/phase · @howeasy
  45. · Evidence links · GitHub search · DeepSeek · pulls · orangeofcarl0-sys/dsh-epistemic-fold · @orangeofcarl0-sys
    Excerpt
    **Campaigns** (model `deepseek-v4.1-flash`, envelope 2,200 requests / 18M input / 1M output / 8h):
  46. · Evidence links · GitHub search · DeepSeek · pulls · openbudgetfun/openbudget · @github-actions[bot]
    Excerpt
    `AgentEndpoint` runs a tool-calling assistant on OpenCode Zen (`deepseek-v4.1-flash`) or the user's own OpenAI-compatible provider, metered at published prices against the included allowance. Bring-your-own keys
  47. · Evidence links · GitHub search · DeepSeek · pulls · nathfroech/map_painting_helper · @renovate[bot]
  48. · Evidence links · GitHub search · DeepSeek · pulls · muni-works/frog · @ecto
  49. · Alex's posts · model family · GitHub search · DeepSeek · pulls · mheci/CachyDeb · @mheci
    Excerpt
    Created with DeepSeek V4.1 Flash 1017 in OpenCode. The CVE scanner was written by a poteto-agent subagent on the same model.
  50. · Evidence links · GitHub search · DeepSeek · pulls · maximhq/bifrost · @sanchitmonga22
    Excerpt
    wally partition, because the model ids name vendors (`wally/deepseek-v4.1-flash` would otherwise be claimed by deepseek).
  51. · Evidence links · GitHub search · DeepSeek · pulls · lidge-jun/opencodex · @andrew05060414
    Excerpt
    the last 30 days included about 15k unpriced `deepseek-v4.1-flash` attempts across four relays. There were also `gemini-4-argon` on Vertex/Cursor/Zed, `cn:`-prefixed Kimi/GLM/MiniMax/Step ids, and `zd/`-prefixed relay
  52. · Evidence links · GitHub search · DeepSeek · pulls · kortix-ai/CompanyOS · @kubet
  53. · Evidence links · GitHub search · DeepSeek · pulls · kixel-ai/rust2 · @kixelated
    Excerpt
    usage limit; resets unknown`, and `opencode run -m opencode-go/deepseek-v4.1-flash` printed `Insufficient balance. Top up at https://opencode.ai/console/wrk_01M28N93AE612R8VKHHW47PJER/billing?credits=1`. The login is restored (`opencode models --standalone` lists the
  54. · Evidence links · GitHub search · DeepSeek · pulls · kixel-ai/rind · @kixelated
  55. · Evidence links · GitHub search · DeepSeek · pulls · kixel-ai/rind · @steelhead99x
  56. · Evidence links · GitHub search · DeepSeek · pulls · dimipaun/fugaro · @fugaro-dev[bot]
    Excerpt
    first-line review round 2: 2 findings (recorded: model deepseek/deepseek-v4.1-flash, at 2f1878d) · review levels: level 1: changes in round 2 on `2f1878d` · model cost
  57. · Evidence links · GitHub search · DeepSeek · pulls · dimipaun/fugaro · @fugaro-dev[bot]
    Excerpt
    first-line review round 1: 3 findings (recorded: model deepseek/deepseek-v4.1-flash, at 13904df) · tests failed on `13904df` · model cost $0.14 of $10 cap · updated
  58. · Evidence links · GitHub search · DeepSeek · pulls · decolua/9router · @shroomgaze
  59. · Evidence links · GitHub search · DeepSeek · pulls · crewAIInc/crewAI · @ats3v
  60. · Evidence links · GitHub search · DeepSeek · pulls · crewAIInc/crewAI · @ats3v
    Excerpt
    model="deepinfra/deepseek-ai/DeepSeek-V4.1-Flash", # DeepInfra honours the vLLM option
  61. · Evidence links · GitHub search · DeepSeek · pulls · blackghostosint/mailbox-plus-website · @blackghostosint
    Excerpt
    the identical gate model + rubric (nous / deepseek-v4.1-flash) run with the endpoint timeout extended to 300s. No revision rounds used.
  62. · Evidence links · GitHub search · DeepSeek · pulls · awesome-dsh-plugin/awesome-dsh-plugin · @Vhow123
  63. · Evidence links · GitHub search · DeepSeek · pulls · avleen/felonybench.ai · @github-actions[bot]
  64. · Evidence links · GitHub search · DeepSeek · pulls · antiwork/gumroad · @gumclaw
  65. · Evidence links · GitHub search · DeepSeek · pulls · anomalyco/models.dev · @models-dev-automation[bot]
  66. · Evidence links · GitHub search · DeepSeek · pulls · amalgamator-ai/amalgamator · @alexm-redhat
    Excerpt
    security fixes, one run per case, same model (DeepSeek-V4.1-Flash), sanitized repos. Baseline = the current phase ladder; new = fused pre-RED + fused resolution. Complexity
  67. · Evidence links · GitHub search · DeepSeek · pulls · aloschilov/pyasc-skill-stack · @aloschilov
    Excerpt
    returning a module. W38 and W39 use fresh DeepSeek-V4.1-Flash generations. W38 selected a signed-int16 table, but counted column chunks for a row slab and chose
  68. · Evidence links · GitHub search · DeepSeek · pulls · adobe/skills · @renovate[bot]
  69. · Evidence links · GitHub search · DeepSeek · pulls · TouchAI-org/TouchAI · @hiqiancheng
    Excerpt
    - Model(s) used: `deepseek-v4.1-flash` (Magpie provider)
  70. · Evidence links · GitHub search · DeepSeek · pulls · TouchAI-org/TouchAI · @hiqiancheng
    Excerpt
    - Model(s) used: deepseek-v4.1-flash - Tool(s) used: Codex CLI via T3 Code harness
  71. · Evidence links · GitHub search · DeepSeek · pulls · TouchAI-org/TouchAI · @hiqiancheng
    Excerpt
    - Model(s) used: GPT-5 via Codex, deepseek-v4.1-flash. - Tool(s) used: Codex/T3 Code and DeepSeek implementation/review workers.
  72. · Evidence links · GitHub search · DeepSeek · pulls · TouchAI-org/TouchAI · @hiqiancheng
    Excerpt
    - Model(s) used: GPT-5 via Codex, deepseek-v4.1-flash. - Tool(s) used: Codex/T3 Code and DeepSeek implementation/review workers.
  73. · Evidence links · GitHub search · DeepSeek · pulls · ThreeNativeHQ/threenative-engine · @jonit-dev
  74. · Evidence links · GitHub search · DeepSeek · pulls · Tencent/LoopForge · @Robert0622
  75. · Evidence links · GitHub search · DeepSeek · pulls · SreeSharvesh/loc-arena · @ryanlail
  76. · Evidence links · GitHub search · DeepSeek · pulls · Significant-Gravitas/AutoGPT · @HhH666-ux
  77. · Evidence links · GitHub search · DeepSeek · pulls · RareFormLabs/vue3-gettext · @renovate[bot]
  78. · Evidence links · GitHub search · DeepSeek · pulls · NixOS/nixpkgs · @mtul0729
    Excerpt
    (model DeepSeek-V4.1-Flash).
  79. · Evidence links · GitHub search · DeepSeek · pulls · NixOS/nixpkgs · @starcraft66
  80. · Evidence links · GitHub search · DeepSeek · pulls · NVIDIA/hyperflare-inference · @jetnet
  81. · Evidence links · GitHub search · DeepSeek · pulls · NVIDIA/hyperflare-inference · @javieratomic
  82. · Evidence links · GitHub search · DeepSeek · pulls · NVIDIA/TensorRT-LLM · @moraxu
  83. · Evidence links · GitHub search · DeepSeek · pulls · MaMissaoui/Finglass · @MaMissaoui
  84. · Evidence links · GitHub search · DeepSeek · pulls · MaMissaoui/Finglass · @MaMissaoui
  85. · Evidence links · GitHub search · DeepSeek · pulls · KarpelesLab/oxinfer · @MagicalTux
    Excerpt
    - DeepSeek V4.1 Flash: catalog entries, the model table row and its section of the DeepSeek V4 page.
  86. · Evidence links · GitHub search · DeepSeek · pulls · BashfulBits/city-meeting-podcasts · @BashfulBits
  87. · Evidence links · DeepSeek · Research & News articles
  88. · Evidence links · DeepSeek · updates
    Excerpt
    API changes DeepSeek V4.1 Flash is now available on the DeepSeek API with native multimodal support
  89. · Alex's posts · Alex Getman
    Excerpt
    Right now, you can test 5 models for free in Cline: Gemini 3.8 Flash MiniMax M3.1 👀 MiMo-V2.6-Flash DeepSeek V4.1 Flash Muse Spark 1.3
  90. · Alex's posts · Alex Getman
    Excerpt
    🚨 DeepSeek V4.1 Flash is out • New architecture: it only uses the part of the model it actually needs, making it faster and cheaper. • New KV cache: long conversations and agent workflows are noticeably cheaper to run. • On coding and agent tasks, it’s already close to Opus 5 and GPT-5.6 Sol. • And all of that at 15x lower cost.

Signals 2

Specifications

DeepSeek V4.1 Flash is a model from DeepSeek with open weights, released on 10 Sep 2026. The context window is 1,048,576 tokens.

Context window
1,048,576
Input
image · text
Output
text
Weights
Open

Questions

What is the context window of DeepSeek V4.1 Flash?

The context window is 1,048,576 tokens.

Are the weights of DeepSeek V4.1 Flash publicly available?

Yes. DeepSeek publishes the weights.

When was DeepSeek V4.1 Flash released?

On 10 Sep 2026.