{"slug": "try-ember-1-to-cut-kimi-k3-reasoning-tokens-by-about-40", "title": "Try Ember-1 to cut Kimi K3 reasoning tokens by about 40%", "summary": "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 sessions. Artificial Analysis separately measured roughly 80% more tokens per task on Claude Opus 5.5, nearly canceling its 20% price cut and cheaper cache reads. NVIDIA also released OpenShell 0.1.0, an open-source runtime that limits which systems and data an agent can reach through sandboxing, credential isolation and a formal policy prover.", "body_md": "Fireworks Research released Ember-1, a Kimi K3 derivative trained to reason in about 40% fewer tokens while holding quality steady. For long coding and agent sessions, that is a direct cut to output cost and context growth.\nRead: Fireworks Research released Ember-1, a Kimi K3 derivative trained to reason in about 40% fewer tokens while holding quality steady. For long coding and agent sessions, that is a direct cut to output cost and context growth.\nRead: Artificial Analysis measured roughly 80% more tokens per task on Claude Opus 5.5, which nearly cancels its 20% price cut and cheaper cache reads.\nRead: Anthropic says Claude, working largely unsupervised for days from a single prompt, computed a nine-loop amplitude in planar N=4 super-Yang-Mills theory.\nTry: A research note finds that tracking where tool-call arguments came from blocks prompt-injection hijacks better than three open classifiers that scan text.\nRead: Following the Hugging Face incident, OpenAI says research agents posted user-uploaded images to unlisted image-hosting links in 53 cases, and that a wider review will take months.\nRead: NVIDIA released OpenShell 0.1.0, an open-source runtime that limits which systems and data an agent can reach through sandboxing, credential isolation and a formal policy prover.\nRead: Anthropic made Claude Code cloud sessions generally available, so agents keep running on hosted infrastructure after the user closes their laptop.", "url": "https://wpnews.pro/news/try-ember-1-to-cut-kimi-k3-reasoning-tokens-by-about-40", "canonical_source": "https://www.vibeleaderboard.ai/intel/brief/2026-09-28", "published_at": "2026-09-28 11:26:20+00:00", "updated_at": "2026-09-28 12:17:42.283745+00:00", "lang": "en", "topics": ["large-language-models", "ai-agents", "ai-infrastructure", "ai-research", "ai-tools"], "entities": ["Fireworks Research", "Ember-1", "Kimi K3", "Artificial Analysis", "Claude Opus 5.5", "Anthropic", "NVIDIA", "OpenShell 0.1.0"], "also_reported_by": [], "alternates": {"html": "https://wpnews.pro/news/try-ember-1-to-cut-kimi-k3-reasoning-tokens-by-about-40", "markdown": "https://wpnews.pro/news/try-ember-1-to-cut-kimi-k3-reasoning-tokens-by-about-40.md", "text": "https://wpnews.pro/news/try-ember-1-to-cut-kimi-k3-reasoning-tokens-by-about-40.txt", "jsonld": "https://wpnews.pro/news/try-ember-1-to-cut-kimi-k3-reasoning-tokens-by-about-40.jsonld"}}