LLMap Intelligence Price Index (LIPI)
The lowest price at which anyone can buy a given level of intelligence, daily, from our append-only price history across every tracked provider.
What this says right now
derived · as of Oct 1The market moved down across the board over 65 days, with the workhorse tier down 82% and the frontier down 80%, so a given budget now reaches much further up the range. The frontier premium over the GPT-4-class floor fell from 7.1 times to 1.8 times, meaning the tiers have moved closer together. Every floor is now set by an open-weight model, so buyers should look first at open-weight options when stretching a budget.
read from Sep 30 — the series below runs to Oct 1
Every tier is cheaper than at the start of this window — the workhorse floor fell fastest, down 82% in 66 days. Frontier intelligence costs 1.8× the GPT-4-class floor, down from 7.1× at the start.
Every price floor on this index — workhorse to frontier — is currently set by a model with open weights. To match the open floor's price at the same bar, the cheapest closed-model offer costs 6.3× more at the frontier and 4.6× more at the workhorse tier.
6 bars, 4 prices today: Ling-2.6-flash clears both the workhorse and capable tier bars; GLM 5.3 Flash clears both the near-frontier and frontier bars. Where one model clears two bars, nothing cheaper sits between them — so read these as 4 observations, not 6. The histories still differ: each bar reached today's price by its own route.
Changes cover the 66 days of price history the index holds so far. What moved is above; why the market moved is your call to make.
The repricings behind it
largest moves · past 7 daysFrom our price history — a lab repricing its own list comes first, then the largest moves. The full feed is on News.
Cheapest price today · pick a tier to chart it
moves over the 66 days the index holds
Frontier floor
The cheapest price anyone charged each day for a model within 50 Elo of the top-rated model that day. Per million tokens, 3 parts input to 1 part output.
Also held the floorGemini 3.5 Flash Lite Aug 6–9 · DeepSeek V4.1 Flash Sep 26–29 · GLM 5.3 Flash Sep 30–Oct 1
Each line is the lowest price anyone charged that day for a model clearing that tier's intelligence bar, per million tokens, counting three input tokens for every output token. A falling line means that level of intelligence got cheaper to buy. A step is a repricing, or a new model taking over as the cheapest way to reach the bar; the largest repricings are listed above.
| Tier | Bar | Today | Cheapest via |
|---|---|---|---|
| Workhorse | Elo ≥ 1250 | $0.015/M | Ling-2.6-flash on OpenRouter |
| Capable tier | Elo ≥ 1300 | $0.015/M | Ling-2.6-flash on OpenRouter |
| GPT-4-class | Elo ≥ 1350 | $0.065/M | gpt-oss-120b on CoreWeave |
| Strong reasoner | Elo ≥ 1400 | $0.086/M | Gemma 4 26B A4B on Darkbloom |
| Near-frontier | Elo ≥ 1450 | $0.12/M | GLM 5.3 Flash on DeepInfra |
| Frontier | ≤50 Elo off #1 · Elo ≥ 1459 today | $0.12/M | GLM 5.3 Flash on DeepInfra |
Source: LLMap Intelligence Price Index (LIPI v1), llmap.ai/insights/lipi, as of 2026-10-01<iframe src="https://llmap.ai/insights/lipi/embed" width="820" height="480" style="border:0;max-width:100%;aspect-ratio:820/480" title="LLMap insight: lipi" loading="lazy"></iframe>