Preview build — figures marked “✓” or in green are sourced (benchmarks by Artificial Analysis); treat everything else (value scores, “Est.” figures, unmarked prices and changes) as illustrative. Methodology

Corrections

Found a wrong number?

Tell us and we will fix it in public. Every number here is meant to be checked — each price links to its source, and the whole dataset is downloadable.

Report an error

Email corrections@modelpricebook.com with the model, what we show, and what the provider's page says. A link to the provider page is enough — no form, no account.

How corrections work

  • Response: we aim to reply within 2 business days and to fix confirmed errors in the next daily data run.
  • Verification:we re-read the provider's own pricing page and record the quote as evidence — the provider page outranks every aggregator feed.
  • While disputed: a contested number is marked under review rather than silently edited, and the conflicting values stay on the record. See the methodology.
  • After the fix: the correction is listed below with before and after values. We do not quietly overwrite history.
  • If it cannot be sourced: a model whose price no provider publishes is retired from the catalogue on record, not guessed at.

Corrections made

Every price on the siteAug 1, 2026

The preview launched with illustrative sample prices. All 31 list prices were replaced with sourced values cross-checked across public feeds and confirmed against provider pricing pages.

Benchmark scoresAug 1, 2026

Synthetic scores labelled with real benchmark names (“Arena ELO”, “MOS”) were removed and replaced with Artificial Analysis measurements. Models with no sourced measurement now show “—” instead of a number.

Qwen3 235B (Together) output priceAug 1, 2026

Two feeds disagreed 10× ($0.60 vs $6.00). Together's own pricing page confirmed $0.60; the $6.00 figure is their fine-tuning rate. The dissenting feed is recorded as outvoted rather than deleted.

Kling 2.5 (fal.ai) priceAug 2, 2026

Listed at $0.35 per second. fal.ai's own figure is $0.35 per 5-second clip — $0.07 per second. Corrected, a 5× error.

Wan 2.6, Seedance 1.5, Veo 3.1 Fast, Kling 3.0 pricesAug 2, 2026

Each corrected against the provider's own page: Wan 2.6 $0.05 → $0.10 (the old figure was the Flash variant), Seedance 1.5 $0.20 → $0.052, Veo 3.1 Fast $0.15 → $0.10, Kling 3.0 $0.10 → $0.168.

Google Cloud TTS WaveNet and Amazon Polly Neural pricesAug 2, 2026

WaveNet $0.016 → $0.004 per 1K characters (Google moved WaveNet onto the Standard SKU); Polly Neural $0.004 → $0.016 (the old figure was Polly Standard's rate).

Retired from the catalogue

Removed because no provider or feed publishes a price for them — listed so the removal is auditable.

DeepSeek R1DeepSeekAug 1, 2026

no longer on DeepSeek's official price list (superseded by the V4 lineup)

DeepSeek V3.2DeepSeekAug 1, 2026

no longer on DeepSeek's official price list (superseded by the V4 lineup)

Mistral Medium 3MistralAug 1, 2026

superseded by Mistral Medium 3.5 on Mistral's price page; the 3.0 rate is no longer published

Llama 4 MaverickFireworksAug 1, 2026

moved off Fireworks serverless per-token pricing (on-demand GPU deployment only)

Llama 4 70BTogether / Fireworks / GroqAug 2, 2026

no such model exists in any connected feed or provider catalogue (the Llama 4 line ships as Scout/Maverick 17B variants) — unsourceable

DeepSeek R1GroqAug 2, 2026

Groq's catalogue does not list DeepSeek R1 in any connected feed — unsourceable

Qwen 3 235BFireworksAug 2, 2026

Fireworks lists no Qwen3-235B serverless offering in any feed or on its pricing pages — unsourceable

GPT-5.2 miniOpenAIAug 2, 2026

absent from all connected feeds and OpenAI's pricing page under any plausible identifier — unsourceable

Mistral Small 3MistralAug 2, 2026

Mistral's price list carries Small 3.2 and Small 4 only; the 3.0 rate is no longer published — unsourceable