#llm-pricing

Inception's Mercury 2.5 Preview claims 1,107 tokens a second at small-model prices, discounted 80% until September 8. What a diffusion LLM changes for agents.

Identical dollar-per-million pricing does not mean identical cost: tokenizers split the same text into different counts. A vocabulary reference table.

DeepSeek shipped a separate experimental V4 Flash vision model on August 21 at text prices, on a peak and off-peak clock that doubles the rate.

OpenAI cut GPT-5.6 Sol to $4 in and $20 out per million tokens on August 21. The same Vercel changelog window added a $0.10/GB-month storage line.

Price-change notice is a contract term, not a courtesy, and the periods range from a fortnight to none at all. What to check before committing model spend.

Crossing a context threshold can reprice an entire request, not just the tokens above the line. Where the cliffs sit and how to keep prompts underneath.

Time-of-day LLM pricing is arriving, and a cheaper window is not the same as a cheaper bill. How the announced schedules and peak hours actually work.

Gemini 3.6 and 3.7 Flash both bill $0.75/$3.75 until December 31, with $1.50/$7.50 posted for January. Budget at the standard rate, not the promo.

Google, OpenAI and Anthropic all cut batch pricing 50 percent, while Grok 4.6 gets no batch discount at all. What the cheap lane actually costs you.

Mistral's Aug 11 launch pins inference to the EU or US for 10% more, adds a 99.5% SLA tier at 75% more, and hosts Z.ai's GLM-5.2 alongside its own.

Gemini 3.7 Flash ships at $0.75/$3.75 per M through Dec 31, 2026 — but Google cut 3.6 Flash to that same intro rate, so half-price needs an asterisk.

Qwen3.8-Max ships as a 2.4T-parameter MoE with 95B active and a 1M context. What the vendor benchmark table shows, and what was only promised at launch.

One week repriced the cheap end of the model market. A surface-labelled table of what Luna, Qwen3.7 Flash and DeepSeek V4 Flash now cost, and what that buys.

Meta's Muse Spark 1.1 and SpaceXAI's Grok 4.5 launched a day apart — two cheap, agentic value models. We compare price, context, tool use and coding.

Meta enters the paid model-API market with Muse Spark 1.1 — a $1.25/$4.25, 1M-context agent model that tops tool-use benchmarks but trails on pure coding.

Grok 4.5, Opus 4.8, and GPT-5.5 compared on benchmarks, cost, context, and reliability — an honest, source-checked read on which model wins which job.

SpaceXAI shipped Grok 4.5 on July 8: co-trained with Cursor, $2/$6 per Mtok, 4.2x fewer tokens than Opus 4.8. Where it fits an agency stack.

Dev shops quote $8K-$500K for the same 'simple agent' with no math. Our itemized 2026 index: stated-assumption build and monthly run costs you can recompute.

Anthropic shipped Claude Sonnet 5 on June 30, 2026. Its most agentic Sonnet yet lands near Opus 4.8 on coding and knowledge work at roughly half the price.

Uber burned a year's AI budget in four months; Microsoft cut Claude Code. A 4-gate playbook that right-sizes model spend and cuts AI bills 60 to 80 percent.

We compare Claude Opus 4.8 and GPT-5.5 on coding, agents, reasoning, and real cost — including where GPT-5.5 still wins and which model fits which job.

Forecast monthly LLM token spend by client tier, workflow, model mix. Downloadable forecasting model + reconciliation cadence for agency CFOs.

Side-by-side input, output, cached, and batch pricing for 30 frontier and open-weight models across 12 providers. Updated April 2026 with 200+ price points.

Xiaomi's MiMo-V2-Pro has 1T+ parameters with 42B active, ranking #8 worldwide. $1/$3 per million tokens. Complete model guide and deployment.

GPT-5.4 Nano at $0.20 per million input tokens is OpenAI's cheapest model. API-only, designed for classification, extraction, and subagents.

Qwen 3.5 medium series: Flash, 35B-A3B, 122B-A10B, and 27B. Benchmarks vs GPT-5 mini and Claude Sonnet 4.5, pricing from $0.10/M tokens.