# Models

[https://timesof.ai/category/models](https://timesof.ai/category/models)

Model releases, benchmarks, capability jumps, and lab research milestones.

## Coverage

- [Meta Releases Muse Wonder, Its First Frontier Model, and Artificial Analysis Puts It Two Points Off the Top](https://timesof.ai/2026/10/meta-muse-wonder-first-frontier-model) — Meta's Muse Wonder and Wonder Full launch at $4/$20 per million tokens. Artificial Analysis scores Wonder Full 56 vs Opus 5.5's 58; safety report pending.
- [OpenAI Puts GPT-6.1 Sol on Ultrafast at $12/$60 per Million Tokens, Selling Up to 8x Speed Across the API, Codex and ChatGPT Work](https://timesof.ai/2026/10/openai-ultrafast-gpt-6-1-sol) — OpenAI rolls out Ultrafast for GPT-6.1 Sol: up to 8x Sol Standard speed at $12/$60 per million tokens, open to all API users; Pro 500 and Enterprise in Codex.
- [Claude Haiku 5.5 Matches GPT-6 Luna's Price on Short Prompts — and Sonnet 5.5's Cache Reads Get Halved](https://timesof.ai/2026/10/anthropic-claude-haiku-5-5-sonnet-cache-cut) — Haiku 5.5: $0.10/$0.50 under 100k tokens, matching GPT-6 Luna. AA index 43 vs Luna 38 at max effort. Sonnet 5.5 cache reads halved to $0.10.
- [Mistral Puts 1T-Parameter Large 4 on a Paid Preview API; Weights by Month-End, AA Scores 38](https://timesof.ai/2026/10/mistral-large-4-preview-1t-open-weights) — Mistral Large 4 paid API preview: ~1.05T/49B active MoE, weights by month-end. AA Intelligence Index 38, behind MiMo/GLM-5.3/Kimi K3. Licence unconfirmed.
- [Update: Reflection Shows Beam — 501B / 23B Active, Apache 2.0 Weights Later This Month](https://timesof.ai/2026/10/reflection-open-weight-imminent) — Reflection unveils Beam (501B/23B active MoE). Self-reported benches near GLM-5.2; Kimi K3 and GLM-5.3 still ahead. Apache 2.0 weights later this month. Not on AA yet.
- [Aleph Alpha Releases Kolibri: Apache 2.0 German/English MoE, 3.46B Active](https://timesof.ai/2026/10/aleph-alpha-kolibri-open-weight) — Kolibri is Aleph Alpha’s Apache 2.0 German/English MoE: 78.1B total, 3.46B active, up to 1M context. AIME 96.9 is the lab’s harness, not an independent board rank.
- [Microsoft’s MAI-Transcribe-2-Streaming Debuts #1 on Artificial Analysis Streaming STT](https://timesof.ai/2026/10/microsoft-mai-transcribe-2-streaming) — Microsoft launches MAI-Transcribe-2-Streaming at claimed Artificial Analysis #1 streaming STT; Voice 2.1 family same day. Intro $0.54/hr audio.
- [OpenAI and Synopsys Partner on GPT-Synopsys — Pre-GA EDA-Native Chip Agents](https://timesof.ai/2026/09/openai-synopsys-gpt-synopsys) — OpenAI and Synopsys announce multi-year GPT-Synopsys preferred-partner deal with revenue share; early engagements only — no GA, pricing, or benchmarks.
- [Gemini 4 Argon Opens as a Fairwind-Gated Cyber Frontier — Guardrails Come Later](https://timesof.ai/2026/09/google-gemini-4-argon) — Gemini 4 Argon ships Fairwind/gov without cyber guardrails; API and AI Ultra later. Lab DeepSWE 77.9%, AutomationBench 51.3%; no Arena Elo yet.
- [GPT-6.1 Sol Claims Near-Astra Work at a Fifth of Astra’s Token Price](https://timesof.ai/2026/09/openai-gpt-61-sol) — GPT-6.1 Sol keeps Sol’s $2/$10 price and claims near-Astra coding and computer use. Benchmarks are OpenAI’s. No Arena Elo yet.
- [Sonnet 5.5 Matches Opus on the Index — and Burns the Most Tokens AA Has Measured](https://timesof.ai/2026/09/anthropic-claude-sonnet-55) — Sonnet 5.5 hits AA Intelligence Index 56, behind Opus 5.5 at 58, at the same $2/$10 sticker. AA’s max-effort cost per task is higher, not lower.
- [Astra Is the Ceiling. Sol and Luna Are What Agent Loops Will Pay.](https://timesof.ai/2026/09/openai-gpt-6-sol-and-luna) — OpenAI cut Sol and Luna API prices 50% vs GPT-5.6 promo. Astra stays the ceiling; Artificial Analysis still scores Opus 5.5 ahead of Sol.
- [Opus 5.5 Is Anthropic’s Defaultable Fable-Class Model — Andon Cash Does Not Agree](https://timesof.ai/2026/09/anthropic-claude-opus-5-5) — Opus 5.5 is Anthropic’s defaultable Fable-class model at lower cache-read rates. Artificial Analysis agrees ahead of Sol; Andon vending cash does not.
- [Grok 4.7 Holds Grok 4.6 Prices and Moves the Specialty Benches](https://timesof.ai/2026/09/xai-grok-4-7) — Grok 4.7 holds Grok 4.6 prices at $2/$6. Lab leads EEBench and Harvey legal; Andon Vending-Bench puts it behind Sol, ahead of Opus 5.5.
- [Grok Voice Transcribe 2.0 Cuts Short-Phrase Errors; Price Stays Put](https://timesof.ai/2026/09/xai-grok-voice-transcribe-2) — Grok Voice Transcribe 2.0 cuts short-phrase WER from 20.6% to 6.8% at unchanged prices. Default API slug flips; AA streaming rank is lab-cited.
- [Salesforce Owns a CRM Model — Koa on Nemotron, Weights Inside the Trust Boundary](https://timesof.ai/2026/09/salesforce-koa-nvidia-nemotron) — Koa is Salesforce’s CRM reasoning model post-trained from Nemotron 3 Super. Weights stay inside Salesforce; 3× fewer errors is vendor-only.
- [Gemini 3.8 Live Claims the Independent Speech-to-Speech Board](https://timesof.ai/2026/09/google-gemini-3-8-live-extended-thinking) — Gemini 3.8 Live Extended Thinking leads Artificial Analysis S2S Quality Index at 82.6. API $0.005/$0.018 per min; powers Search Live and Gemini Live.
- [Astra Is Critical Cyber — and Comes With a $1B Defender Subsidy](https://timesof.ai/2026/09/openai-gpt-6-astra) — GPT-6 Astra is OpenAI’s first Critical-cyber model, with $1B Daybreak defender subsidies. Lab benches are vendor; Arena/AA not yet claimed here.
- [Muse Spark 1.3 Closes on Fable on Price — Not Yet on Artificial Analysis](https://timesof.ai/2026/09/meta-muse-spark-1-3) — Muse Spark 1.3 cuts tool calls ~20% vs 1.2. Artificial Analysis Intelligence 61 at xhigh; Fable 5.1 max still ahead at 66.
- [Google’s Third Flash in Six Weeks Splits Cyber Into a Gated SKU](https://timesof.ai/2026/09/google-gemini-3-8-flash-cyber-fairwind) — Gemini 3.8 Flash holds $0.75/$3.75 intro pricing through year-end. Flash Cyber is Fairwind-gated for defenders — Google’s third Flash in six weeks.
- [Fable 5.1 Sells Cheaper Cache and Customer-Held Monitors — Mythos Stays Gated](https://timesof.ai/2026/09/anthropic-claude-fable-5-1-mythos) — Fable 5.1 GA and Mythos 5.1 trusted-access; cache reads cut 75% to $0.25/MTok. Enterprise Frontier Safeguards keep monitors in the customer’s cloud.
- [Gemini Omni 1.1 Flash Is a Director’s API — Extend, Interpolate, Upscale](https://timesof.ai/2026/08/google-gemini-omni-1-1-flash) — Gemini Omni 1.1 Flash ships with 40-second scene extension, first/last frames, and 4K upscaling — a director’s API for developers.
- [Gemini 3.5 Transcribe Folds ASR Into the Gemini Name — Chirp 3 Replaced](https://timesof.ai/2026/08/google-gemini-3-5-transcribe) — Gemini 3.5 Transcribe ships with 4.0% streaming WER on Artificial Analysis, 85+ languages, replacing Chirp 3.
- [Qwen3.8-Flash-Next: A 6B-Active 125B MoE Previewing Qwen4 Cheap](https://timesof.ai/2026/08/alibaba-qwen3-8-flash-next) — Alibaba open-sourced Qwen3.8-Flash-Next, a 125B MoE with 6B active parameters previewing Qwen4 — vendor benches only so far.