AI BriefingDeepSeek V4.1 Flash 1017 scores 45 on Artificial Analysis Intelligence Index
·2026.10.19 13:41
Key point
The open-weights MoE model achieves a score of 45, significantly above the median of 19 for similar-sized models, while offering 230.4 tokens/sec output speed.
Details
- DeepSeek V4.1 Flash 1017 (Max) achieves an Artificial Analysis Intelligence Index score of 45, ranking #2 out of 121 models in its class.
- The model features a 552B total parameter MoE architecture with 16B active parameters per token and a 1M token context window.
- Output speed reaches 230.4 tokens per second, compared to a median of 77.9 tokens per second for similar-sized open-weight models.
- Pricing is set at $0.30 per 1M input tokens and $1.20 per 1M output tokens, with a 98% cache discount available.
Model Specifications
- Released in October 2026 under the MIT license with open weights available on Hugging Face.
- Supports text and image input with text output, utilizing a reasoning model architecture.
- Available via five providers with performance metrics based on the first-party DeepSeek API.
Performance and Cost
- Intelligence Index evaluation generated 210M output tokens, indicating high verbosity compared to the 140M median.
- Cost per Intelligence Index task averages $0.26, with a blended rate of $0.18 per 1M tokens.
- Time to first token (TTFT) is 1.08 seconds, lower than the 1.69-second median for comparable models.
This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.
Related
Ant Group Releases Finance-Specialized Open-Weight Model 'Ling-3.0-flash-Fin'Artificial Analysis Releases Capability Indices v1.1… Claude Fable 5.1 Tops All 6 Categories, Grok 4.5 Noted for Cost Efficiency