InferIndex market index
On , the level of the family “Frontier closed models” was $3.50 per million tokens (at a 3:1 input:output mix), across 95 models: 1 model entered and 9 left since the publication of .
Market
-
FRTFrontier closed models · 95 models
Frontier closed
$3.500.0% change on the same modelsModels entered since the previous publication (1) : inception/mercury-3. On 26 October 2026, the level of the family “Frontier closed models” was $3.50 per million tokens (at a 3:1 input:output mix), across 95 models: 1 model entered and 9 left since the publication of 19 October 2026.Models left since the previous publication (9) : anthropic/claude-opus-4, anthropic/claude-opus-4.1, anthropic/claude-sonnet-4, openai/gpt-3.5-turbo, openai/gpt-4, openai/gpt-4-turbo, openai/gpt-4o-2024-05-13, openai/o1, openai/o1-pro
Frontier closed models · 95 models counted -
CMPCompact closed models · 37 models
Compact closed
$0.4630.0% change on the same modelsModels left since the previous publication (3) : openai/gpt-4.1-nano, openai/o3-mini, openai/o4-mini. On 26 October 2026, the level of the family “Compact closed models” was $0.463 per million tokens (at a 3:1 input:output mix), across 37 models: no model entered and 3 left since the publication of 19 October 2026.
Compact closed models · 37 models counted -
CODCode models · 10 models
Code
$0.4630.0% change on the same modelsModels left since the previous publication (4) : openai/gpt-5.1-codex, openai/gpt-5.1-codex-max, openai/gpt-5.1-codex-mini, openai/gpt-5.2-codex. On 26 October 2026, the level of the family “Code models” was $0.463 per million tokens (at a 3:1 input:output mix), across 10 models: no model entered and 4 left since the publication of 19 October 2026.
Code models · 10 models counted -
OWLOpen weights, 200B+ parameters · 61 models
Open weights 200B+
$0.50+11.1% change on the same models
Open weights, 200B+ parameters · 61 models counted -
OWMOpen weights, 40B to 200B · 30 models
Open weights 40–200B
$0.3290.0% change on the same models
Open weights, 40B to 200B · 30 models counted -
OWSOpen weights, under 40B · 56 models
Open weights <40B
$0.1280.0% change on the same models
Open weights, under 40B · 56 models counted
Market and official list prices, by family
| FamilyA group of similar models the market index tracks together, such as closed frontier models or open models under 40 billion parameters. Each family's value is the median price of its models. | MarketBuilt from the offers we collect from providers and resellers, excluding promotions.$ per 1M tokens | Official list pricesBuilt from the list prices on the labs' own pricing pages.$ per 1M tokens |
|---|---|---|
| Frontier closed (Frontier closed models) | $3.5095 models | $4.8139 models |
| Compact closed (Compact closed models) | $0.46337 models | $0.41114 models |
| Code (Code models) | $0.46310 models | No value Fewer than 3 models this week: no value.1 model |
| Open weights 200B+ (Open weights, 200B+ parameters) | $0.5061 models | $1.06 Corrections · Value before correction $1.12. Published 26 October 2026, corrected on 30 October 2026: a repair of our data or calculation, not a market movement. The reason is listed under Corrections on this page.10 models |
| Open weights 40–200B (Open weights, 40B to 200B) | $0.32930 models | $0.225 Corrections. Published 26 October 2026, corrected on 30 October 2026: a repair of our data or calculation, not a market movement. The reason is listed under Corrections on this page.5 models |
| Open weights <40B (Open weights, under 40B) | $0.12856 models | $0.5575 models |
Official list prices: detail by family
Official list prices
-
FRTFrontier closed models · 39 models
Frontier closed
$4.810.0% change on the same modelsModels left since the previous publication (6) : openai/gpt-3.5-turbo, openai/gpt-4, openai/gpt-4-turbo, openai/gpt-4o-2024-05-13, openai/o1, openai/o1-pro. On 26 October 2026, the level of the family “Frontier closed models” was $4.81 per million tokens (at a 3:1 input:output mix), across 39 models: no model entered and 6 left since the publication of 19 October 2026.
Frontier closed models · 39 models counted -
CMPCompact closed models · 14 models
Compact closed
$0.4110.0% change on the same modelsModels left since the previous publication (3) : openai/gpt-4.1-nano, openai/o3-mini, openai/o4-mini. On 26 October 2026, the level of the family “Compact closed models” was $0.411 per million tokens (at a 3:1 input:output mix), across 14 models: no model entered and 3 left since the publication of 19 October 2026.
Compact closed models · 14 models counted -
CODCode models · 1 model
Code
No value Fewer than 3 models this week: no value.
Code models · 1 model counted -
OWLOpen weights, 200B+ parameters · 10 models
Open weights 200B+
+0.1% change on the same models
Open weights, 200B+ parameters · 10 models counted -
OWMOpen weights, 40B to 200B · 5 models
Open weights 40–200B
+0.3% change on the same models
Open weights, 40B to 200B · 5 models counted -
OWSOpen weights, under 40B · 5 models
Open weights <40B
$0.5570.0% change on the same models
Open weights, under 40B · 5 models counted
How it is computed
- Market series: a model's reference price is the median of its three cheapest providers (one price per provider) among offers in force at the measured instant; models with fewer than 3 providers are left out. Promotions, prices in points, stale offers, flex and batch tiers and resales below the lab's list price are excluded, among others. From the 12 October 2026 publication (method 1.2), a discount declared by a provider with no announced end date for 14 days or more counts as a normal price. From the same publication, an offer whose price change we have not yet confirmed at the measured instant is not counted. From the 26 October 2026 publication (method 1.3), a model whose retirement by its own lab has taken effect is left out entirely from both series, all offers included.
- Official series: a model's reference price is the lab's own list price, undiscounted, standard tier, cheapest region. If the lab's pricing page can no longer be read, its last list price counts for 14 days at most; after that, the model leaves the series until the page is read again.
- A family's value is the unweighted median of its models' reference prices. A family with fewer than 3 models has no value that week.
- Two series, never mixed: the market series from the offers we collect, the official series from the labs' list prices.
- Change on the same models compares only the models counted in both publications, so a model entering or leaving a family does not move the figure.
curl https://api.inferindex.dev/index?series=marketCorrectionsOfficial list prices
- Open weights, 200B+ parameters: from $1.12 to $1.06 (value corrected on 30 October 2026).
- Open weights, 40B to 200B: from – to $0.225 (value corrected on 30 October 2026).
Official series, open_large and open_medium: since 20 October 2026, Zhipu's pricing page puts z-ai/glm-4.7, z-ai/glm-4.5-air, z-ai/glm-4.5v and z-ai/glm-4.6v in quoted boxes that our reading skipped, so these four models dropped out of the series although Zhipu had not retired them. No price moved: the page read again on 26 October 2026 shows the same prices as before (fix of 26 October 2026).