Frontrumors: stories https://frontrumors.com/ Frontier Model News and Rumors You Care About. Release cycles, expected dates, leaks, benchmarks and API pricing. en-us Sun, 01 Nov 2026 11:05:00 GMT tools/build_site.mjs Gemini 4 Carbon Is Official, With Restricted Access and No Public Launch Date https://frontrumors.com/news/gemini-4-carbon-restricted-rollout/ https://frontrumors.com/news/gemini-4-carbon-restricted-rollout/ Sun, 01 Nov 2026 11:05:00 GMT Frontrumors Google announced Carbon on October 28 and began a restricted rollout to cyber defenders and trusted partners. It leads Artificial Analysis' current Intelligence Index v5, but it is not a generally available Gemini API release. News Gemini 4 Carbon Google restricted rollout benchmarks Benchmarks Refresh: Carbon Leads AA Index v5; Reasoning Settings Matter https://frontrumors.com/news/artificial-analysis-index-v5-carbon-leads/ https://frontrumors.com/news/artificial-analysis-index-v5-carbon-leads/ Sun, 01 Nov 2026 11:00:00 GMT Frontrumors Our November 1 snapshot uses Artificial Analysis Intelligence Index v5 and preserves the reasoning settings reported by AA. Carbon scores 59, Opus 5.5 58 and GPT-6 Astra 57; Kimi's current 46 should not be compared directly with its older-index 55. News benchmarks Artificial Analysis Gemini 4 Carbon Kimi K3.1 Pro OpenAI Launches GPT-6.1 Terra, a $1/$5 Computer-Use and Vision Tier Below Sol https://frontrumors.com/news/gpt-6-1-terra-computer-use-tier/ https://frontrumors.com/news/gpt-6-1-terra-computer-use-tier/ Tue, 27 Oct 2026 14:55:00 GMT Frontrumors GPT-6.1 Terra is a fast, efficient model built around visual reasoning and computer use. It lists at $1/$5 per million tokens, half the price of GPT-6.1 Sol, with a 1.05M-token context window. News GPT-6.1 Terra OpenAI computer use pricing Anthropic Files S-1 for Proposed IPO, Plans Nasdaq Listing Under 'ANT' https://frontrumors.com/news/anthropic-files-s-1-nasdaq-ant/ https://frontrumors.com/news/anthropic-files-s-1-nasdaq-ant/ Tue, 27 Oct 2026 14:30:00 GMT Frontrumors Anthropic has filed a registration statement on Form S-1 with the SEC for a proposed IPO of its Class A common stock. The share count and price range are not set, and the announcement does not discuss models or products. News Anthropic IPO S-1 Nasdaq Artificial Analysis: Opus 5.5 Holds No. 1 at 58, Sonnet 5.5 and Kimi K3.1 Pro Join the Top Five https://frontrumors.com/news/artificial-analysis-sonnet-5-5-kimi-k3-1-pro-top-five/ https://frontrumors.com/news/artificial-analysis-sonnet-5-5-kimi-k3-1-pro-top-five/ Tue, 27 Oct 2026 14:05:00 GMT Frontrumors We rechecked Artificial Analysis and LMArena on October 27. Claude Opus 5.5 leads on 58, Sonnet 5.5 and Muse Wonder Full follow on 56, and Moonshot's Kimi K3.1 Pro sits on 55. Our Benchmarks page has five new rows. News benchmarks Artificial Analysis LMArena Sonnet 5.5 Kimi K3.1 Pro Moonshot's Kimi K3.1 Pro Lands at $2/$10 With Open Weights Promised by November 4 https://frontrumors.com/news/kimi-k3-1-pro-api-launch-weights-by-nov-4/ https://frontrumors.com/news/kimi-k3-1-pro-api-launch-weights-by-nov-4/ Tue, 27 Oct 2026 13:40:00 GMT Frontrumors Kimi K3.1 Pro is available in Kimi's products and API. Moonshot promises full weights by November 4; the current AA Intelligence Index v5 score is 46, distinct from the older-index 55 reported at publication. News Kimi K3.1 Pro Moonshot open weights pricing Pricing Alert: Gemini 4 Argon Steps Up to $4/$20 on December 1 https://frontrumors.com/news/gemini-4-argon-price-doubles-december-1/ https://frontrumors.com/news/gemini-4-argon-price-doubles-december-1/ Tue, 27 Oct 2026 13:10:00 GMT Frontrumors Google's pricing page lists Gemini 4 Argon at $2/$10 per million tokens through November 30 and $4/$20 from December 1, a month before Gemini 3.8 Flash's own step-up on January 1. News pricing Gemini 4 Argon Gemini 3.8 Flash API Mistral Large 4 Open Weights Expected This Week as End-of-October Deadline Arrives https://frontrumors.com/news/mistral-large-4-weights-expected-this-week/ https://frontrumors.com/news/mistral-large-4-weights-expected-this-week/ Sun, 25 Oct 2026 22:26:00 GMT Frontrumors Mistral promised Large 4 weights by the end of October when the model entered public preview on October 6. With days left in the month, the open-weight release is the nearest expected launch on our calendar. Rumor Mistral Large 4 open weights release date Grok 5 Leak: ~6-Trillion-Parameter MoE Training on Colossus 2, Q1 2027 Window Firming Up https://frontrumors.com/news/grok-5-colossus-2-six-trillion-leak/ https://frontrumors.com/news/grok-5-colossus-2-six-trillion-leak/ Sat, 24 Oct 2026 18:02:00 GMT Frontrumors Grok 5 is in training on Colossus 2, and the numbers circulating, a ~6T-parameter MoE, native video temporal reasoning, ~1.5M-token context, would make it the largest announced model. The spring beta window has slipped to Q1 2027. Leak Grok 5 xAI Colossus 2 leak Review: Claude Opus 5.5 vs GPT-6 Astra, the Autumn Agentic Coding Showdown https://frontrumors.com/news/opus-5-5-vs-gpt-6-astra-coding-showdown/ https://frontrumors.com/news/opus-5-5-vs-gpt-6-astra-coding-showdown/ Thu, 22 Oct 2026 16:40:00 GMT Frontrumors Two September flagships, two philosophies. Opus 5.5 leads its tier on agentic coding with 1M context at $4/$20; GPT-6 Astra counters at $10/$50 with the highest ceiling, if your organization can get access. Review review Claude Opus 5.5 GPT-6 Astra coding benchmarks Gemini 4 Pro Expected Late Q4 as Google Completes the Gemini 4 Family https://frontrumors.com/news/gemini-4-pro-expected-late-q4/ https://frontrumors.com/news/gemini-4-pro-expected-late-q4/ Tue, 20 Oct 2026 21:15:00 GMT Frontrumors Argon was positioned as the first Gemini 4 model. Following Google's yearly flagship cadence, a Pro tier with deeper reasoning, agents and Workspace integration is expected before the year is out. Rumor Gemini 4 Pro Google rumor release date Guide: Which Frontier Model Should You Migrate to in October 2026? https://frontrumors.com/news/which-frontier-model-october-2026-guide/ https://frontrumors.com/news/which-frontier-model-october-2026-guide/ Sun, 18 Oct 2026 17:00:00 GMT Frontrumors Eight model families refreshed inside eight weeks. Our October shortlist: Opus 5.5 for coding agents, Grok 4.7 for value, DeepSeek V4.1-Flash for self-hosting, and three releases worth waiting for. Guide guide migration buyers guide October 2026 API Pricing Alerts: Gemini Flash Doubles January 1, GPT-5.6 Sol Promo Runs Through at Least November 21 https://frontrumors.com/news/api-pricing-alerts-flash-doubling-gpt-promo-ending/ https://frontrumors.com/news/api-pricing-alerts-flash-doubling-gpt-promo-ending/ Fri, 16 Oct 2026 15:30:00 GMT Frontrumors The two most important dates in near-term API pricing: Gemini 3.8 Flash's promo rate doubles to $1.50/$7.50 on New Year's Day, and OpenAI's GPT-5.6 Sol promo pricing is guaranteed only through at least November 21. News pricing deals Gemini 3.8 Flash GPT-5.6 API Gemini 4 Argon Reaches Stable GA, First Model From Google's New Pre-Training Run https://frontrumors.com/news/gemini-4-argon-stable-ga/ https://frontrumors.com/news/gemini-4-argon-stable-ga/ Tue, 13 Oct 2026 19:00:00 GMT Frontrumors Thirteen days after its September 30 announcement, Gemini 4 Argon is now Stable GA, the first Gemini 4 model and the opening move in Google's annual flagship season. News Gemini 4 Argon Google GA release Meta's Muse Wonder Enters Preview at Frontier Level, Level With GPT-6 Astra https://frontrumors.com/news/muse-wonder-preview-frontier-level/ https://frontrumors.com/news/muse-wonder-preview-frontier-level/ Mon, 12 Oct 2026 17:30:00 GMT Frontrumors Muse Wonder (muse-wonder-1.0) is Meta's most capable model ever, previewing at ~52.4 on the Artificial Analysis index. It is preview-only with no published API pricing, and it redraws Meta's roadmap. News Muse Wonder Meta preview frontier Claude Haiku 5.5 Completes Anthropic's Five-Week Model Refresh https://frontrumors.com/news/claude-haiku-5-5-fast-tier/ https://frontrumors.com/news/claude-haiku-5-5-fast-tier/ Wed, 07 Oct 2026 16:00:00 GMT Frontrumors Haiku 5.5 brings the fast tier up to the 5.5 generation, five weeks after Fable 5.1 opened Anthropic's autumn run. The entire Claude line has now been refreshed since September 1. News Claude Haiku 5.5 Anthropic release Mistral Large 4 Enters Public Preview, Open Weights Promised by End of October https://frontrumors.com/news/mistral-large-4-public-preview/ https://frontrumors.com/news/mistral-large-4-public-preview/ Tue, 06 Oct 2026 20:20:00 GMT Frontrumors Mistral's new flagship is available to evaluate today at roughly $1.36/$4.18, but the headline is the promise attached to it: open weights before the month is out. News Mistral Large 4 Mistral preview open weights GPT-6.1 Sol Lands as OpenAI's Balanced Tier in a Three-Wave GPT-6 Rollout https://frontrumors.com/news/gpt-6-1-sol-balanced-tier/ https://frontrumors.com/news/gpt-6-1-sol-balanced-tier/ Tue, 29 Sep 2026 23:45:00 GMT Frontrumors First Astra for limited orgs, then Sol + Luna, now GPT-6.1 Sol as the balanced middle. OpenAI's GPT-6 family is assembling in public through the autumn rollout. News GPT-6.1 Sol OpenAI GPT-6 release Claude Opus 5.5 Ships With 1M Context, Sets the Agentic Coding Pace for the Autumn https://frontrumors.com/news/claude-opus-5-5-agentic-coding/ https://frontrumors.com/news/claude-opus-5-5-agentic-coding/ Tue, 22 Sep 2026 16:15:00 GMT Frontrumors A same-day double feature, Grok 4.7 arrived September 21, but Opus 5.5 at $4/$20 with 1M context is the release engineering teams will benchmark their agents against this quarter. News Claude Opus 5.5 Anthropic coding release DeepSeek V4.1-Flash Ships With MIT Open Weights on Day One at $0.15/$0.60 https://frontrumors.com/news/deepseek-v4-1-flash-mit-weights/ https://frontrumors.com/news/deepseek-v4-1-flash-mit-weights/ Thu, 10 Sep 2026 18:30:00 GMT Frontrumors DeepSeek's September flagship did not bother with an API-only window: V4.1-Flash arrived September 10 with MIT weights the same day and hosted pricing around $0.15/$0.60, keeping DeepSeek at the aggressive end of open-weight pricing. News DeepSeek V4.1-Flash DeepSeek open weights release Qwen3.8-Max Tops Alibaba's Line as the Quarterly Max Cadence Rolls On https://frontrumors.com/news/qwen3-8-max-tops-alibaba-line/ https://frontrumors.com/news/qwen3-8-max-tops-alibaba-line/ Mon, 03 Aug 2026 17:00:00 GMT Frontrumors Alibaba's Qwen3.8-Max arrived August 3 with a 1M context at $2/$6, continuing a Max cadence that has run roughly quarterly through 2026. It remains the multilingual and Alibaba Cloud pick in our guide. News Qwen3.8-Max Alibaba Qwen release