Pricing trends
How LLM API pricing is evolving — price drops, provider competitiveness, and market movements over time. Built from twice-daily snapshots the scan never overwrites.
State of LLM pricing
market overview — computed from the current scanModels tracked
245
current models — deprecated and preview tiers excluded
Providers
40
every price read from the provider's own pricing page
Cheapest paid · blended
$0.041 /Mtok
Granite 4.0 H Micro — IBM
Most expensive · blended
$262.50 /Mtok
o1-pro — OpenAI
How prices are moving
movement — aggregated over a 90-day window, ranked by sizeRepricing activity over the 90 days to . For the full chronological log of every launch, price move and context change — plus JSON and Atom feeds — see the Newsroom.
Repricings in 90d
15
tracked list-price changes inside the window
Price drops
8
blended price went down — cheaper than before
Price increases
7
the moves that show up on invoices
Median move size
60%
half of the moves in the window were smaller than this
Biggest moves
Ranked by size of the move, not by date — the question the chronological feed can't answer.
| Model | Move | Blended $/1M | In $/1M | Out $/1M | Size | Date |
|---|---|---|---|---|---|---|
| DeepSeek V4 FlashDeepSeek | increase | $0.175 → $0.660 | $0.140 → $0.440 | $0.280 → $1.32 | 277.1% | |
| DeepSeek V4 ProDeepSeek | increase | $0.544 → $1.98 | $0.435 → $1.32 | $0.870 → $3.96 | 264.1% | |
| Solar Pro 4Upstage | increase | $0.158 → $0.525 | $0.090 → $0.300 | $0.360 → $1.20 | 233.3% | |
| Solar Pro 4Upstage | increase | $0.053 → $0.158 | $0.030 → $0.090 | $0.120 → $0.360 | 200% | |
| GLM-5.3-FlashZ.AI | increase | $0.119 → $0.238 | $0.075 → $0.150 | $0.250 → $0.500 | 100% | |
| GPT-5.6 LunaOpenAI | drop | $2.25 → $0.450 | $1.00 → $0.200 | $6.00 → $1.20 | 80% | |
| ParetoUnbiased | drop | $3.75 → $1.40 | $2.50 → $0.800 | $7.50 → $3.20 | 62.7% | |
| Qwen3.7-MaxTogether | increase | $1.88 → $3.00 | $1.25 → $2.00 | $3.75 → $6.00 | 60% |
Who reprices most
Providers by number of tracked price changes in the window — a proxy for how actively each one competes on price.
Provider competitiveness
median blended $/Mtok per providerMedian blended cost per provider — lower means more competitive on price. The count is how many current models each provider lists.
What we're tracking
coverage — how the history is builtModel Price Watch takes twice-daily snapshots of every model's pricing across all major LLM providers. Each snapshot records input price, output price, cached input price, context window, and status — so we capture not just what changed, but when and by how much.
Currently tracking 245 models across 40 providers. Snapshots are captured automatically twice a day by our pricing pipeline. Historical data is never overwritten — each snapshot is append-only, so every recorded price point is preserved.
Dig deeper into the token economy:
- Full price-change feed — every recorded input/output move across all providers, newest first
- Marketwide price index — the overall direction of LLM API costs, with the intelligence-cost curve
- Per-model price history — input/output/cached price points over time for every model