#AI Models

OpenAI shuts down gpt-5, gpt-5-mini, gpt-5-nano, gpt-5-pro, o3 and o3-pro on December 11, 2026. The snapshots, the named replacements and what the switch costs.

NVIDIA's Nemotron 3.5 Transcribe, released October 28, against Whisper large-v3 and Parakeet TDT 0.6B v3 on licence, languages, size, hardware and WER.

Mistral released Large 4 weights on October 29 and lists $0.70 input and $2.10 output per million tokens. Direct prices compared with DeepSeek, Kimi and GLM.

Google announced Gemini 4 Carbon on October 28, with access phased to cyber defenders via Fairwind. Who qualifies, the announced price and what to do meanwhile.

Recursive says its Eureka Machine helped build Flint, a 40B open coding model, in under four months. What is claimed, what is shown, and how to get access.

Ideogram Vector 1.0 generates editable SVG files from a prompt or an image, priced per file. What editable means, what it costs and what to check first.

Mistral's Voxtral 3 is four separate speech models in Public Preview. What each does, what it costs per minute or character, and what preview leaves unsettled.

MiniMax M3.1 Flash costs $0.15 per million input tokens, reads images and video, and targets agents and coding. Benchmarks, limits and who should try it.

Now that Gemini 4 Argon is generally available, how it compares with GPT-6.1 Sol and Claude Opus 5.5 on price, output limits, shared benchmarks and fit.

Gemini 4 Argon is now GA in the Gemini API at $2/$10 until November 30. What changes when you switch: output caps, caching, sampling and timeouts.

OpenAI will remove GPT-4o, GPT-4o mini, GPT-4.1 and GPT-4.1 mini from its API on January 14, 2027. What breaks, what to switch to and what it costs.

October 2026 AI model releases with dated prices, licences and access notes. Terra, Genie 4, LFM3 and speech/document updates through October 28.

Gemini 4 Argon launches at $2/$10 per million tokens, rising to $4/$20 later. Who can use it now, what Google's benchmarks show and how to prepare.

GPT-6.1 Sol approaches Astra on key agent tasks at $2/$10 per million tokens. Compare pricing, cache costs, benchmark limits and the case for switching.

Four publishers now list decision models on OpenRouter that return probabilities for typed questions, not prose, at free to $0.05 per million input tokens.

Claude Sonnet 5.5 rejects thinking disabled, forced tool choice and three more settings. The exact errors and fixes, grouped by the model you run today.

Anthropic's charts give 40 score and cost points for Sonnet 5.5 and Opus 5.5. Up to $2 a task Sonnet wins three of four; at $5, Opus wins all four.

Claude Sonnet 5.5 keeps Sonnet 5's $2/$10 price, scores 70.6% on Terminal-Bench 4.0 and adds cyber fallbacks. What the 30% saving is measured on.

ElevenLabs' Eleven v4 tops Artificial Analysis' voice arena and v4 Turbo claims ~100ms. The two latency figures, the vendor tests and the v4 price.

Chinese models reached 57–67% of OpenRouter tokens in reported regional groups. What CNBC’s September figures measure, their limits and what buyers can learn.

92 text-to-speech models ranked by blind listener vote, with the price per million characters, open-weight licences and the job each leader fits.

Fireworks tuned Kimi K3 into Ember-1 and says it matches K3 with 35 to 50% shorter reasoning at the same price. The rows it loses, and the preview caveat.

Gemini 3.8 Flash TTS is generally available with voice cloning from a 30-second sample. The promotional price ends December 31 and doubles on January 1, 2027.

A census of 21 training-cost figures labs published themselves, 2022 to 2026. Six give a dollar amount, none is audited, and each leaves something out.

Xiaomi says six days of RL on MiMo-V2.6 cost about $850,000 for Flash and $2.62 million for Pro. What the figure covers, what it leaves out, and the prices.

Artificial Analysis ran GPT-6 Sol and Claude Opus 5.5 on the same tests. Sol is cheaper up to a point; Opus 5.5 at its default outscores Sol at max.

GPT-6 Sol costs $2/$10 and Luna $0.10/$0.50 per million tokens, half GPT-5.6's price. What OpenAI's own charts show about scores and effort.

Opus 5.5 outscores Fable 5.1 on every coding benchmark Anthropic published, at 40% of the token price. When Fable 5.1 still earns its place.

Grok 4.7 lists at under a third of Opus 5.5's output price, yet costs twice as much per task in Artificial Analysis tests. Where Muse Spark 1.3 fits.

Claude Opus 5.5 lists at 40% of GPT-6 Astra's price with no long-context surcharge. Where each leads, and where the vendors' benchmarks disagree.

Claude Opus 5.5 costs $4/$20 per million tokens, defaults to medium effort and breaks four Opus 5 integrations. What the 40% saving assumes.

Six signals that a model is shipping, five checkable and one not, from stealth tests to alias flips, with the API field for each and three launches scored.

xAI released Grok 4.7 on September 21, 2026 at Grok 4.6's $2/$6 list price. The vendor-run benchmark table, the route price you are billed, three questions.

16 models in the open-model conversation, checked against the licence file, OSI's list and what else was released: weights, data, recipes, evaluation harness.

A 4B model trained for $1,200 cut Postgres query latency 44.7% on a standard benchmark. The trait that made it work, and a routing table for your own tasks.

Qwen's September 18 model takes audio and video natively with 1M context and, by Qwen's own method, cuts per-hour audio cost 98% versus its predecessor.

TypeSafe AI's Jev returns typed values with probabilities, not text, and claims 70 to 500 ms responses. What the vendor figures mean and how to test them.

Compare Fugu Max and Ultra v2 pricing, long-context charges and review costs. Use a workload test to decide when additional reasoning justifies its price.

A dated ledger of AI model releases in September 2026, each row checked against a vendor primary where one exists, with price, context and what it replaces.

On August 12, DeepSeek's pricing page listed V4-Pro-0813 with no changelog entry and a homepage saying nothing had changed. GA was announced a day later.

xAI's Grok 4.6 keeps Grok 4.5's headline rates, but the $2/$6 tier only holds below a 200K-token prompt: cross it and the whole request reprices.

Alibaba previewed Qwen3.8-Max at WAIC Shanghai, claiming 2.4 trillion parameters and second only to Fable 5 — yet shipped zero benchmarks to back it.

Twelve AI models launched in a single week of March 2026 from OpenAI, Google, Mistral, xAI, and more. Developer guide to capabilities, pricing, and selection.

xAI releases Grok 4.20 with 2M token context, lowest hallucination rate at 78%, and 60% lower pricing. Three API variants for reasoning and multi-agent use.

GPT-5.4 ships three variants: Standard, Thinking, and Pro. Native computer use, 1M context, tool search, and 33% fewer factual errors. Complete guide.

Sarvam AI releases 105B and 30B open-source reasoning models under Apache 2.0, trained entirely in India. 128K context window powers the Indus AI assistant.

OpenAI teased GPT-5.4 the same day as GPT-5.3 Instant launch. Rumored 2M context window, enhanced reasoning, and what it means for the AI model roadmap.

OpenAI releases GPT-5.3 Instant with 26.8% fewer hallucinations, 400K context, and anti-cringe tone overhaul. Complete benchmarks, pricing, and migration guide.

Google launches Gemini 3.1 Flash-Lite at $0.25 per million input tokens. 2.5x faster, tops 6 benchmarks. Complete pricing and performance comparison guide.

OpenAI's GPT-5.3-Codex brings 25% faster inference and major Terminal-Bench and OSWorld gains. Full benchmarks, access details, and migration guide.

Claude Opus 4.6 brings a 1M token context window (beta at launch), adaptive thinking and 128K output. Guide to its benchmarks, pricing and API changes.

OpenAI launches GPT-5.2 and GPT-5.2-Codex optimized for agentic coding as GPT-4o retires February 13, 2026. Complete migration and implementation guide.

Master Grok 4.1: EQ-Bench #1 ranking, 65% hallucination reduction, Fast API access, xAI benchmarks, and comparison with GPT-5.2 and Claude Opus 4.5.

GPT-5.2 guide: Instant, Thinking and Pro tiers, 70.9% GDPval wins or ties, 30% fewer error-containing answers, $1.75/$14 API pricing and integration tips.

Master GPT-5.1 Instant and Thinking models. 8 personalities, 2-3x faster. Complete guide with API and ChatGPT integration.

Deploy GLM 4.6 with current Z.ai, OpenRouter, vLLM, and SGLang guidance covering endpoints, pricing, MIT licensing, model IDs, and production caveats.

Qwen3 model-family guide refreshed with 2026 caveats for Qwen3.6-Max-Preview, Coder, Thinking models, and deployment strategies.

Historical Kimi K2-0905 release guide refreshed for 2026 context, with agentic coding, MoE architecture, deployment, and newer Kimi caveats.

DeepSeek-V3.1 guide preserved as historical 2025 context, with 2026 caveats for newer DeepSeek V3.2 and V4 reasoning models.

Historical GPT-5 launch guide refreshed with 2026 caveats for GPT-5.4 and GPT-5.5, capabilities, pricing, benchmarks, and API access.

Gemini 2.5 Flash, Pro, and Deep Think guide refreshed with 2026 caveats for Gemini 3.1, pricing, benchmarks, and model selection.