The Fireworks AI blog index shows a dense, recent run of platform and model news in July 2026, led by two 7/26 posts: "Make Kimi K3 Yours: LoRA Training on Fireworks" and "Fireworks Nexus: Drop-in Open Frontier Intelligence for Teams with Budgets." Earlier in the month, Fireworks announced its Series D and "$1B ARR" on The same index also documents the day-zero availability of multiple frontier open models on Fireworks inference — GLM 5.2 (6/16), GLM 5.2 Fast (6/30), NVIDIA Nemotron 3 Ultra (6/4), Kimi K2.7 Code (6/12), Qwen 3.7 Plus (6/12), and the MiniMax M3 live post (6/12) — plus operational changes like the 7/1 prepaid billing m

Learn how you can access low latency, high throughput inferencing for open models and performance-optimized deployment of custom models with Fireworks AI on Microsoft Foundry.

Fireworks Ai: SaaS platform for developers and enterprises to run, fine-tune, and deploy LLMs and image models with fast inference and low latency.

The Fireworks AI changelog documents several notable platform updates through August 2026. On August 25, SSO documentation was updated to confirm that IdP-initiated SAML is now supported as an opt-in feature via --enable-idp-initiated-sso, correcting earlier guidance that incorrectly stated only SP-initiated login was Earlier entries show the addition of new models to the Model Library, including GLM 5.2 (June 15), and Kimi K2.7 Code, MiniMax M3, and Qwen 3.7 Plus (June 12). Multiple serverless deprecations occurred in June, including Kimi K2.5, Qwen 3.6 Plus, and MiniMax M2.5, with recommended migration paths to newer variants. On

IREN's 7/20/2026 press release announces $2.8 billion in new multi-year AI Cloud contracts and raises its 2026 AI Cloud ARR target from $3.7bn to more than $4bn. Fireworks AI is named directly as a contracted customer alongside Microsoft, NVIDIA, Perplexity, Figure AI, Together AI, Fluidstack, Fal AI, and Hume AI, with The deal is operationally significant for Fireworks: it independently corroborates that Fireworks is securing substantial multi-year GPU/compute capacity from a vertically integrated AI Cloud provider during a period of supply tightness. IREN reports scaling from roughly 3MW of self-built AI Cloud capacity a year earli