DeepSeek released DeepSeek V4.1 Flash, a 552 billion parameter mixture-of-experts model under an MIT license that activates only 8 billion parameters during prefill and 16 billion during decode, cutti…
DeepSeek released V4.1 Flash on September 10, 2026, cutting cache-hit costs to $0.003 per token during off-peak hours from the $0.022 charged for the outgoing V4-Pro, a 77-80% price reduction, and rai…
DeepSeek V4-Flash costs about 98% less than OpenAI's GPT-5.5 for heavy workloads, with monthly costs dropping to $12.60 versus around $210 for the same 30M/30M token workload. DeepSeek's pricing is $0…
Frontier AI models are activating far fewer parameters per token than three years ago — Zhipu's GLM-5.2 activates about 40 billion, Alibaba's Qwen3.5 runs 17 billion active, and DeepSeek V4-Flash gets…
Reasoning models are deliberately trading world knowledge for reasoning skill, with GLM-5.2 scoring 99.2% on AIME 2026 using about 40 billion active parameters per token, while factual recall remains …
DeepSeek V4, released as a preview on April 24, 2026, and fully available on July 20, 2026, offers a 1-million-token context window and pricing of $0.435 per million input tokens and $0.87 per million…
DeepSeek is raising API pricing for its V4 model family by up to 1,100%, with new peak and off-peak rates effective August 16, as demand strains capacity. The increases eliminate DeepSeek's price adva…
Floatboat's V4-Flash model, evaluated by its own Floatboat Evaluation Harness, achieved top rankings on five third-party benchmarks, including first place on Toolathlon (79.62) and third on BrowseComp…
DeepSeek's official API documentation lists V4-Flash at $0.14 per million cache-miss input tokens, $0.28 per million output tokens, and $0.0028 per million cache-hit input tokens, making it unusually …
A 1.6 trillion parameter model like DeepSeek V4 can run on a laptop by keeping 97% of its parameters asleep and streaming experts from disk, according to a technical analysis that includes a complete …
An engineer ran an agent session that shipped three pull requests to pnp/sp-dev-fx-webparts in 75 minutes of active work, yet 40% of the engineer's messages were corrections. The agent repeatedly viol…