DeepSeek says new downloadable AI beats pricier rivals at coding for about 20 cents a job
DeepSeek's own tests show a smaller model beating bigger rivals at coding, but outside tests are missing and China may limit such releases.
DeepSeek has released a new AI model called DeepSeek-V4.1-Flash-1017, The Decoder reports. It works on DeepSeek's website and app. The model files are also posted on Hugging Face and ModelScope, so anyone can download them.
DeepSeek says the model is faster and writes shorter answers. It says the model is better at building full software projects. In its own tests, it beats larger models like Claude Opus 5 and GPT 5.6-Sol. On one coding test it scored 77.0 percent at just under $0.20 per task. The older version scored 74.2 percent at around $0.50. DeepSeek's chart puts rivals at $2 to over $10 per task, and none match its score.
These results come from DeepSeek itself. Outside tests aren't out yet. DeepSeek is also opening its own search engine to developers, at $0.75 per 1,000 uses. China's government reportedly plans to tighten rules on publishing open models, or even block releases.
In my opinion, the pressure here is on price. If DeepSeek's numbers hold up, rivals charging $2 to $10 per task must explain the gap, and builders get a cheap model to download. The catch is that China's reported plans could limit future releases like this one.
The facts
- DeepSeek released DeepSeek-V4.1-Flash-1017 on its platforms and as an open-weight model on Hugging Face and ModelScope.
- In DeepSeek's own benchmarks, the model beats larger rivals including Claude Opus 5, GPT 5.6-Sol, Kimi K3 and GLM 5.3 in several coding and agent categories.
- On the DeepSWE test, the model hits 77.0 percent at an average cost of just under $0.20 per task.
- The predecessor scored 74.2 percent at around $0.50 per run.
- Rival models in DeepSeek's chart cost between $2 and over $10 per task, and none match the smaller DeepSeek model's accuracy.
- Independent benchmarks of the model haven't come out yet.
- DeepSeek's Search and Fetch APIs cost $0.75 per 1,000 calls per endpoint, and are free through certain tools until October 31, 2026.
- The Chinese government reportedly plans to tighten rules on publishing open-weight models or block their release entirely.
Sources
- The Decoder: Deepseek V4.1 Flash 1017 beats larger rivals in agentic tasks while once again massively cutting per-task costs
Published
More in New models
AmazonNew modelsAyobi’s pickReflection AI's Beam, a free-to-download coding AI, is now available on Amazon's cloud
Reflection AI released Beam, an AI built to write and fix software over long jobs. Anyone can download and change it, and Amazon now offers it to run on its cloud.
New modelsOfficial releaseAyobi’s pickvia Hugging FaceIBM releases AI models that chat, make quick calls and transcribe speech in one
IBM says its new Granite Switch 4.1 models handle conversation, yes-or-no style judgments and speech-to-text from one online connection, using the memory of one model.
New modelsvia The DecoderBlack Forest Labs releases FLUX 3 Image Klein, an image AI for home computers
The new model makes and edits pictures up to 4K size. Black Forest Labs says it runs on an NVIDIA graphics card with 8 GB of memory.