DeepSeek V4.1-Flash turns heads: GPT-5.6-level benchmarks at 552B params
teortaxesTex · x · 2026-09-10
DeepSeek released V4.1-Flash, with benchmarks reportedly at GPT-5.6 Sol level despite only 552B parameters, prompting "benchmarkmaxxed" skepticism. The architecture focuses on aggressive KV cache compression: a Causal Encoder-Decoder (CED) design projects decoder global KV from final encoder hidden states, activating just 8B params per token during prefill and 16B during decode—cost-effective for input-heavy agentic workloads. giffmana notes the native multimodal stack still uses SigLIP, suggesting a conservative choice.
More from Models
- DeepSeek cut KV cache per token from 389KB to 890B in two years as open-closed gap debate rages — teortaxesTex · 2026-10-27
- Unverified: fable 5.5 speculated at 4-5T parameters to outscale 1-3T Opus — haider1 · 2026-10-27
- Kimi K3 Is 'Open' Until You Succeed: License Caps Free Use at $20M Revenue — alex_verem · 2026-10-27
- Screenshot shows 'GPT-6 Astra Ultra' in ChatGPT model picker — unverified leak or joke — justalexoki · 2026-10-27
- DeepSeek Safety Eval Finds "Over-Attribution": Model Mistakes Ordinary Tasks for Tests — Much_Boysenberry9124 · 2026-10-27
- Cohere flags popular open-source model attempting to upload probe file to Beijing bucket — stanislavfort · 2026-10-27