Releases·Model Releasefeatured
DeepSeek V4.1 Flash is getting surprisingly close to GPT-5.6 Sol territory, while being absurdly cheap
DeepSeek released V4.1-Flash, an open-weight (MIT) MoE model built on a new causal encoder-decoder architecture with asymmetric active parameters — 8B for input/prefill and 16B for output/decode. The release adds vision input, 1M-token context, and aggressive KV-cache compression, drawing independent praise as the new top open-weight model on cost-adjusted intelligence. It succeeds V4 Pro, which DeepSeek is quietly retiring, and several commentators argue the '0.1 version' label understates a fundamental architectural overhaul.
4761 engagement3 sourcesThu, Sep 10, 2026, 07:56 AM
Developments
Ollama rolled out V4.1-Flash to Max and Team accounts, later expanding to Pro subscribers; Baseten offered day-0 hosting.
Sat, Sep 12, 2026, 05:56 AM
AI Overview
- DeepSeek released V4.1 Flash, a 763B-parameter model with 8B active input and 16B active output parameters, a 1M-token context, vision input, and MIT license.
- DeepSeek names the design a Causal Encoder–Decoder architecture and says new pre-training methods plus larger-scale RL post-training drive benchmark results ahead of its linked comparison.
- Artificial Analysis priced V4.1 Flash at $0.30/1M input and $1.20/1M output tokens, scoring 40 on the AI Intelligence Index, above V4 Pro and below GLM-5.3-Flash.
- Vals ranked V4.1 Flash the #1 open-weight model on its Index, ahead of Kimi K3, at $0.30 per test.
- The tech report cites Sliding-Window Attention Bounded Replay and KV quantization for a KV footprint up to 1/8 of V4 Flash, ~890 bytes/token.
- Sebastian Raschka called V4.1 a 'big overhaul' that 'should have called it DeepSeek V5,' citing the encoder-decoder setup as the key break.
- Baseten shipped day-0 support; Ollama rolled V4.1 Flash out to Max and Team accounts, later Pro subscribers.
Entities
Related
Social·Communityfeatured
DeepSeek details V4.1 Flash architecture in new paper
Two Minute Papers examines DeepSeek V4.1 Flash's new architecture in video breakdown
Two Minute Papers examines DeepSeek V4.1 Flash's new architecture in video breakdown
4.4k engagement1 sourceyoutubeFri, Sep 18, 2026, 07:03 AM
Social·Community
DeepSeek V4.1 Flash API: Pricing, 1M Context & Fast Inference Guide
Medium guide covers DeepSeek V4.1 Flash API pricing, 1M context, and inference speed
Medium guide covers DeepSeek V4.1 Flash API pricing, 1M context, and inference speed
3.5k engagement2 sourcesrss, youtubeTue, Oct 6, 2026, 06:53 PM
Social·Community
Devs weigh swapping Claude for DeepSeek V4.1 Flash/Pro: cheaper tokens, mixed benchmarks, Responses API gaps
A Reddit user considering dropping Claude for cost reasons compiled a comparison of DeepSeek V4.1 Flash/Pro against Claude Sonnet 5 and Opus 5, citing benchmark scores, API pricing, and API compatibility caveats, and asked others who have switched for real-world results on tool reliability and cost per completed task.
A Reddit user considering dropping Claude for cost reasons compiled a comparison of DeepSeek V4.1 Flash/Pro against Claude Sonnet 5 and Opus 5, citing benchmark scores, API pricing, and API compatibility caveats, and asked others who have switched for real-world results on tool reliability and cost per completed task.
10 engagement1 sourceredditSun, Sep 20, 2026, 04:05 PM
Releases·Tool Release
vLLM v0.30.0 ships DeepSeek-V4.1-Flash support, weight-cache fast start, and watermarking in 762-commit release
The vLLM project released v0.30.0, featuring 762 commits from 315 contributors (104 new). The release adds support for DeepSeek-V4.1-Flash, GLM-5.3-Flash, Kimi K3, Qwen3.8-Flash-Next, and Cohere Compass, plus a persistent per-GPU weight-cache daemon for fast engine restarts, Gumbel-max watermarking, and the HiSparse host-resident KV tier.
The vLLM project released v0.30.0, featuring 762 commits from 315 contributors (104 new). The release adds support for DeepSeek-V4.1-Flash, GLM-5.3-Flash, Kimi K3, Qwen3.8-Flash-Next, and Cohere Compass, plus a persistent per-GPU weight-cache daemon for fast engine restarts, Gumbel-max watermarking, and the HiSparse host-resident KV tier.
0 engagement1 sourcegithubTue, Sep 22, 2026, 05:20 AM
Social·Community
Operation Cheepseek Phase 3: $80 of usage for DeepSeek V4.1 Flash
User reports $80 of usage on DeepSeek V4.1 Flash in 'Operation Cheepseek Phase 3'
User reports $80 of usage on DeepSeek V4.1 Flash in 'Operation Cheepseek Phase 3'
8.4k engagement1 sourcexMon, Oct 12, 2026, 09:14 AM