Frontier input-token floor

The Frontier input-token floor is USD 0.50 per 1M input tokens on 2026-10-24 (n=10 eligible models), set by llm/official/mistral/mistral-large. Definition: The lowest standard input price, in USD per million tokens, among the current flagship text models of the providers in the frontier basket.

Computed from the SKU state at the end of each UTC day (latest value: 2026-10-24; today's value is recomputed until the day ends). The constituents of the latest value are listed below; each shows when it was verified and links to its source.

Change over this window: 0.0% from 2026-10-20 to 2026-10-24 (like for like, on the constituents both days share).

Download this chart: SVG · PNG · CSV · JSON

Constituents of the latest value

Basket status

Each basket model is decided by its first-party price (llm/official/…). CONFIRMED: community sources agree. DISPUTED: they differ, first-party wins. PROVISIONAL: only the lab's LiteLLM row exists (community cross-check, no first-party price read yet). UNCONFIRMED: no price, not in the series.

Disagreements

Models where a community source differs from the deciding value by more than the threshold. Shown, not resolved.

Details of the latest value

Side

input

Min coverage

3

Excluded retired count

0

First party models

10

Provisional models

0

Disagreement count

3

Disagreement threshold pct

5

Excluded provisional count

0

Basket date

2026-10-24

Basis

first-party prices only

SKU ID

llm/official/mistral/mistral-large

Basket median

2

Basket ceiling

10

stale constituents left out

0

coverage changed

no

{
 "chain": {
  "broken": false,
  "change_pct": 0.0,
  "index": 100.0,
  "matched": 10,
  "prev_day": "2026-10-23",
  "ratio": 1.0
 },
 "constituents": {
  "anthropic/claude-fable-5-1": 10.0,
  "anthropic/claude-opus-5-5": 4.0,
  "anthropic/claude-sonnet-5-5": 2.0,
  "deepseek/deepseek-v4-pro": 1.32,
  "gemini/gemini-3.8-flash": 0.75,
  "gemini/gemini-4-argon": 2.0,
  "mistral/mistral-large": 0.5,
  "openai/gpt-6-astra": 10.0,
  "openai/gpt-6.1-sol": 2.0,
  "xai/grok-4.7": 2.0
 },
 "status_counts": {
  "CONFIRMED": 7,
  "DISPUTED": 3
 }
}

Definition

The lowest standard input price, in USD per million tokens, among the current flagship text models of the providers in the frontier basket.

Inclusion rule

hand-reviewed dated basket (FRONTIER in code); a model is eligible when a first-party price exists (llm/official/<provider>/<model> decides, matched on provider plus a normalized model key, lower case with '.' and '-' alike: CONFIRMED when community values agree within 1e-9 USD/token, DISPUTED when they differ and first-party wins). Only when a basket model has no official SKU is the LiteLLM row of the lab itself read and the model labelled PROVISIONAL (basis recorded per model in basis_per_model): it stays in the status table but never sets or counts toward the floor (excluded_provisional lists it). Models with neither are UNCONFIRMED and not eligible. Standard text input, lowest tier, USD per 1M tokens. The number of models where models.dev, OpenRouter or LiteLLM differ from the deciding value by more than 5% is published (disagreement_count). A basket model whose stated retirement date (LiteLLM deprecation_date or a first-party deprecation page) is on or before the computed day is excluded and listed in excluded_retired.

When it cannot be computed: Not computed: that day shows no number and states the reason; the last full value stays shown with its date. A source whose last confirmed snapshot is older than 3x its cadence (6h sources 18 h, 12h sources 36 h, 1d sources 72 h) drops out; below minimum coverage the series is not computed and says 'not verified: <source> stale since <date>'.

Minimum coverage: 3 eligible models. Below that the series is not computed and says so instead of showing a number.

Sources feeding it

llm/official/<provider>/<model> (first-party pricing pages), litellm (fallback/cross-check), modelsdev (cross-check), openrouter (cross-check)

Deliberately excluded

• batch, cached, priority and flex prices
• off-peak tariffs
• models without a price

What can break it

• a page redesign breaks the regex recipe (the model becomes UNCONFIRMED)
• model renames
• a provider moves a model to another tier
• basket edits (dated basis changes)
• if the cheapest provider is DeepSeek the floor is mostly DeepSeek; the status table shows the rest

Frontier set (fixed list, reviewed by hand)

llm/anthropic/claude-fable-5-1, llm/anthropic/claude-opus-5-5, llm/anthropic/claude-sonnet-5-5, llm/deepseek/deepseek-v4-pro, llm/gemini/gemini-3.8-flash, llm/gemini/gemini-4-argon, llm/mistral/mistral-large, llm/openai/gpt-6-astra, llm/openai/gpt-6.1-sol, llm/xai/grok-4.7

A model that is in the list but not (yet) in the data simply does not contribute; n shows how many did. Until first-party price pages are parsed, these prices come from LiteLLM, a community cross-check.

Daily values, last 90 days

Days with a value since 2026-10-20 (days without one are left out of the data files): CSV · JSON.

API: JSON · CSV

InfraRabbit (2026). Frontier input-token floor [data as of 2026-10-24]. https://infrarabbit.com/index/frontier-input-floor (accessed 2026-10-24). List prices, not transacted prices. CC BY 4.0.

@misc{infrarabbit_index_frontier_input_floor_2026,
  author       = {{InfraRabbit}},
  title        = {Frontier input-token floor},
  year         = {2026},
  howpublished = {\url{https://infrarabbit.com/index/frontier-input-floor}},
  note         = {Data as of 2026-10-24; accessed 2026-10-24; list prices, not transacted prices; CC BY 4.0}
}