GitHub deprecated four models across all GitHub Copilot experiences on October 2, 2026: Gemini 3.5 Flash, Gemini 3.6 Flash, Kimi K2.7 Code, and Claude Opus 4.7. GitHub recommends Gemini 3.8 Flash for …
DepthFirst launched an AI-native continuous cyber defense product that scans code, dependencies, infrastructure and live environments, and published a benchmark comparing 34 model and thinking-level c…
OpenAI said in a report published Wednesday that it shut down a coordinated model-distillation campaign that spiked to 16,000 requests from more than 4,000 users on July 24 and 25, attributing "a core…
D-Matrix co-founder and CTO Sudeep Bhoja told Let's Data Science that Raptor's early 3D-DRAM silicon uses 0.37 picojoules per bit moved, roughly five to eight times better than the 2 to 3 picojoules p…
Moonshot AI's Kimi K3 is now available in OpenAI's enterprise Codex channel through US inference provider Baseten, letting enterprise customers count use of the Chinese model against their existing Op…
CAISI's assessment of Z.ai's GLM-5.3, released August 14, 2026 with weights published two weeks later, found the model scored 40.4% (74/183) on SEC-Bench Pro, 61.1% (9.8/16) on ExploitBench, 9.4% (47/…
A developer published a practical guide to integrating the Kimi K3 chat completions API through Ace Data Cloud's endpoint, covering request fields, response parsing, streaming, and multi-turn dialogue…
Running an open-weight model at the level of Opus 5.5 requires roughly 610 GB of VRAM plus RAM for a maximally quantized Kimi K3, while the full-precision model would need 1600 GB, according to a Leve…
A Peking University research team led by Haoran Ye trained the behavior of a specialized agent harness into Qwen3.5-9B via a method called harness distillation, raising the small model's average task …
A wave of open-weight model releases this summer — including Zhipu's GLM-5.3, Moonshot AI's Kimi K3, DeepSeek's V4 Pro and V4.1 Flash, and Xiaomi's MiMo-V2.6 — narrowed the gap with top proprietary mo…
Cognition cut Devin pricing by 30-40% in Fusion and Normal modes, 15-20% in Ultra mode, and up to 70% for Devin Review, while its Fusion harness scored 68.8 on the Extended portion of the FrontierCode…
A developer asked 13 AI models from 10 companies to describe how they imagine themselves, collecting 116 answers that were stripped of model names and coded blind. None of the 116 answers chose a huma…
Fireworks Research released Ember-1, a Kimi K3 derivative trained to reason in about 40% fewer tokens while holding quality steady, cutting output cost and context growth for long coding and agent ses…
Fireworks AI released Ember-1, a post-trained version of Moonshot AI's open-weight Kimi K3 that delivers Kimi K3's quality with about 40% fewer tokens, available only through the Fireworks serverless …
Fireworks AI released Ember-1 on September 22, 2026, a fine-tuned variant of Moonshot AI's Kimi K3 that cut reasoning token consumption 71.3% and total token spend 39% in live production A/B tests on …
Anthropic's Claude Opus 5.5 won Roberto Nickson's five-model 3D-printed bridge challenge, holding an estimated 130 lb versus 26.5 lb for Meta's Muse Spark 1.3 and an estimated 17.5 lb for OpenAI's GPT…
Fireworks AI says its Ember-1 model, built from Moonshot AI's Kimi K3, produces comparable answers with about 40% fewer tokens, according to a September 27 X post by co-founder Dmytro Dzhulgakov promo…
Fireworks post-trained Kimi K3 to create Ember-1, a model that uses roughly 40% fewer tokens while maintaining comparable quality across Fireworks' evaluations, according to the company's announcement…
Fireworks Research released Ember-1, a specialized model built on Kimi K3 that delivers Kimi K3's quality using 40% fewer tokens, developed through more than 50 training experiments and over 200 evalu…
Anthropic's Claude Opus 5.5 and Moonshot AI's Kimi K3 each independently found the same novel link-cable vulnerability in Pokémon Emerald and built a working worm that turns a victim's game into Tetri…