benchrHow this log works
The log follows three rules. Append-only: entries are not edited or removed after publication. Same-day verification: a number enters the log only on a day benchr confirms it on a provider pricing page or announcement, not from a search snippet or third-party table. Source attached: every entry links to the record used for verification.
The append-only log begins on June 1, 2026 and does not backfill earlier prices. Older documented changes appear below as separate context, with their own sources.
The log so far
| Logged | Model | Event | Input | Output |
|---|---|---|---|---|
| 2026-09-03 | GPT-6 Astra | New model | $10 | $50 |
| 2026-09-02 | Gemini 3.8 Flash | New model | $0.75 | $3.75 |
| 2026-09-01 | Claude Fable 5.1 | New model | $10 | $50 |
| 2026-08-31 | Claude Sonnet 5 | Price change | $2 | $10 |
| 2026-08-28 | DeepSeek V4-Pro | Price change | $1.32 | $3.96 |
| 2026-08-28 | DeepSeek V4-Flash | Price change | $0.44 | $1.32 |
| 2026-08-27 | Qwen3.8-Flash | New model | $0.15 | $0.47 |
| 2026-08-24 | GPT-5.6 Sol | Price change | $4 | $20 |
| 2026-08-24 | Gemini 3.6 Flash | Price change | $0.75 | $3.75 |
| 2026-08-21 | Kimi K3 | New model | $3 | $15 |
| 2026-08-21 | Grok 4.6 | New model | $2 | $6 |
| 2026-08-21 | GLM-5.3 | New model | $1.4 | $4.4 |
| 2026-08-21 | Gemini 3.7 Flash | New model | $0.75 | $3.75 |
| 2026-08-03 | Qwen3.8-Max | New model | $2 | $6 |
| 2026-07-13 | MiniMax M3 | Model added | $0.3 | $1.2 |
| 2026-07-13 | Grok 4.5 | Model added | $2 | $6 |
| 2026-07-13 | GPT-5.6 Sol | Context and pricing re-verified | $5 | $30 |
| 2026-07-13 | GLM-5.2 | Model added | $1.4 | $4.4 |
| 2026-07-09 | GPT-5.6 | New model | $5 | $30 |
| 2026-07-08 | gpt-5-6 | Correction | — | — |
| 2026-07-08 | gemini-3-5-pro | Correction | — | — |
| 2026-07-08 | claude-sonnet-5 | Correction | — | — |
| 2026-07-03 | Claude Sonnet 5 | Price change | $2 | $10 |
| 2026-07-01 | Claude Sonnet 5 | New model | $4 | $20 |
| 2026-06-30 | Gemini 3.5 Pro | New model | $2.5 | $15 |
| 2026-06-10 | GPT-5.4 | New model | $2.5 | $15 |
| 2026-06-10 | Claude Fable 5 | New model | $10 | $50 |
| 2026-06-01 | Mistral Large 3 | New model | $0.5 | $1.5 |
| 2026-06-01 | Llama 4 Scout | New model | — | — |
| 2026-06-01 | Llama 4 Maverick | New model | — | — |
| 2026-06-01 | Kimi K2.6 | New model | $0.95 | $4 |
| 2026-06-01 | Grok 4.3 | New model | $1.25 | $2.5 |
| 2026-06-01 | GPT-5 Mini | New model | $0.25 | $2 |
| 2026-06-01 | GPT-5.5 | New model | $5 | $30 |
| 2026-06-01 | GPT-5 | New model | $1.25 | $10 |
| 2026-06-01 | Gemini 3.5 Flash | New model | $1.5 | $9 |
| 2026-06-01 | DeepSeek V4-Pro | New model | $0.435 | $0.87 |
| 2026-06-01 | DeepSeek V4-Flash | New model | $0.14 | $0.28 |
| 2026-06-01 | Claude Opus 4.8 | New model | $5 | $25 |
Most entries are baselines: the verified starting point each future change gets measured against. That's how an honest price database begins: you can't detect a change without a trustworthy "before."
Documented moves from before the log
These predate June 1, 2026, so they live outside the append-only dataset, but each is documented by the provider's own announcements and worth keeping in view:
The Opus lane fell two-thirds. Claude Opus 4 and 4.1 billed $15/$75 per million tokens through 2025. Since Opus 4.5 (November 2025), the lane bills $5/$25 — and both old models retire this summer, making the cut universal.
GPT-4o halved within months. The May 2024 launch snapshot billed $5/$15; later 2024 snapshots billed $2.50/$10. The original snapshot, still billing 2024 prices, shuts down October 23, 2026.
Reasoning got cheap. o1 launched December 2024 at $15/$60, and o1-pro hit $150/$600 in March 2025, OpenAI's priciest model ever. Today GPT-5.5 reasons better at $5/$30.
DeepSeek reset the recorded floor twice, in both directions. The June 1, 2026 log captured V4-Pro at $0.435/$0.87 after a reported 75% reduction on api-docs.deepseek.com. On August 16, 2026 DeepSeek raised the whole V4 line and split it into peak and off-peak halves, so V4-Pro reads $1.32/$3.96 peak and $0.66/$1.98 off-peak, and V4-Flash $0.44/$1.32 and $0.22/$0.66. Peak hours are 01:00–04:00 and 06:00–10:00 UTC, Monday to Friday. Those figures describe dated rate cards; they can change, so the current provider page remains the source for a new purchase or budget.
And the increases
Published base rates are only one source of cost changes. Google's October 16 cutoff moves Gemini 2.5 Flash traffic to a model billing 5× more per input token. xAI re-pointed legacy Grok API slugs to Grok 4.3 billing, changing the price of unchanged integrations. Gemini's over-200K-token surcharge also raised effective long-context rates for some workloads. For that reason, this log records retirements, routing changes, and migration effects alongside rate-card updates.
What gets logged
Six event types match the dataset schema: price_change, new_model, benchmark_update, deprecation_announced, sunset, and context_change. Pricing means published list rates for input, output, and, where offered, cached input. Promotional and negotiated enterprise rates stay out. Open-weight records with no license charge are logged as $0-license entries; that is not a $0 infrastructure claim. Verified changes also appear in the site changelog and affected pricing pages.