Buyer tools
LLM API & GPU cost calculators
Price it yourself. Three checks on the site's current data: what a month of LLM API traffic would cost, where a GPU rental quote sits among comparable posted list prices, and what Azure listed for 1- and 3-year reservations. Everything is calculated in your browser; your configuration is kept in the page address so you can share it.
LLM API
Monthly API bill
Enter a month of traffic. Each model is priced at its standard list price, with a cached-input or batch price only where the table records one for that model, and at its long-context price once your prompts reach its tier. Where the table records a long-context price but not where it starts, prompts that reach the earliest tier it does record are ranked at the higher price.
Example month: three input tokens for every output token, the mix behind the blended price on /inference, so this default ranks models the way that column does. Cached-input prices are recorded for 36 of 37 models. GPT-5.5 Pro has none, so its cache hits are billed at the full input rate. Batch prices are recorded for 31 of 37 models. 6 models have none, so their batch traffic is billed at standard rates.
Context windows are not checked: the table does not record them, so every model is priced at the prompt size you enter, even one longer than it accepts.
- 1
GPT-6 Luna · OpenAI
input $3.00 · output $5.00 · checked 2026-10-15
Note: ≤272K prompt tokens; above 272K $0.20/$0.75
$8.00a month
- 2
Claude Haiku 5.5 · Anthropic
input $3.00 · output $5.00 · checked 2026-10-15
Note: ≤100K prompt tokens; above 100K $0.50/$2.50
$8.00a month
- 3
GPT-5.6 Luna · OpenAI
input $6.00 · output $12.00 · checked 2026-10-15
Note: ≤272K prompt tokens; above 272K $0.40/$1.80
$18.00a month
- 4
GPT-5.4 nano · OpenAI
input $6.00 · output $12.50 · checked 2026-10-15
Note: Deprecated; shutdown Apr 1, 2027 (replacement GPT-6 Luna)
$18.50a month
- 5
DeepSeek V4.1 Flash · DeepSeek
input $9.00 · output $12.00 · checked 2026-10-15
Note: Peak-hour cache-miss input; off-peak $0.15/$0.60; deepseek-flash alias
$21.00a month
- 6
Gemini 3.1 Flash-Lite · Google
input $7.50 · output $15.00 · checked 2026-10-15
Note: Text/image/video input; audio input $0.50
$22.50a month
- 7
Gemini 3.5 Flash-Lite · Google
input $9.00 · output $25.00 · checked 2026-10-15
$34.00a month
- 8
Gemini 3.8 Flash · Google
input $22.50 · output $37.50 · checked 2026-10-15
Note: Introductory through Dec 31, 2026; $1.50/$7.50 from Jan 1, 2027
$60.00a month
- 9
Gemini 3.6 Flash · Google
input $22.50 · output $37.50 · checked 2026-10-15
$60.00a month
- 10
Grok 4.3 · xAI
input $37.50 · output $25.00 · checked 2026-10-15
Note: Below 200K prompt tokens; at/above 200K $2.50/$5
$62.50a month
- 11
GPT-5.4 mini · OpenAI
input $22.50 · output $45.00 · checked 2026-10-15
$67.50a month
- 12
DeepSeek V4 Pro 0813 · DeepSeek
input $39.60 · output $39.60 · checked 2026-10-15
Note: Peak-hour cache-miss input; off-peak $0.66/$1.98
$79.20a month
- 13
Claude Haiku 4.5 · Anthropic
input $30.00 · output $50.00 · checked 2026-10-15
$80.00a month
- 14
Grok 4.7 · xAI
input $60.00 · output $60.00 · checked 2026-10-15
Note: Below 200K prompt tokens; at/above 200K $4/$12
$120.00a month
- 15
Grok 4.6 · xAI
input $60.00 · output $60.00 · checked 2026-10-15
Note: Below 200K prompt tokens; at/above 200K $4/$12
$120.00a month
- 16
Grok 4.5 · xAI
input $60.00 · output $60.00 · checked 2026-10-15
Note: Below 200K prompt tokens; at/above 200K $4/$12
$120.00a month
- 17
GPT-6.1 Sol · OpenAI
input $60.00 · output $100.00 · checked 2026-10-15
Note: ≤272K prompt tokens; above 272K $4/$15
$160.00a month
- 18
GPT-6 Sol · OpenAI
input $60.00 · output $100.00 · checked 2026-10-15
Note: ≤272K prompt tokens; above 272K $4/$15
$160.00a month
- 19
Claude Sonnet 5.5 · Anthropic
input $60.00 · output $100.00 · checked 2026-10-15
$160.00a month
- 20
Claude Sonnet 5 · Anthropic
input $60.00 · output $100.00 · checked 2026-10-15
Note: Intro $2/$10 made standard Aug 10, 2026; Sep 1 rise to $3/$15 cancelled
$160.00a month
- 21
Gemini 4 Argon · Google
input $60.00 · output $100.00 · checked 2026-10-15
Note: Introductory through Nov 30, 2026; $4/$20 from Dec 1, 2026
$160.00a month
- 22
GPT-5.6 Terra · OpenAI
input $60.00 · output $120.00 · checked 2026-10-15
Note: ≤272K prompt tokens; above 272K $4/$18
$180.00a month
- 23
Gemini 3.1 Pro Preview · Google
input $60.00 · output $120.00 · checked 2026-10-15
Note: ≤200K prompt tokens; above 200K $4/$18
$180.00a month
- 24
GPT-5.4 · OpenAI
input $75.00 · output $150.00 · checked 2026-10-15
Note: ≤272K prompt tokens; above 272K $5/$22.50
$225.00a month
- 25
Claude Sonnet 4.6 · Anthropic
input $90.00 · output $150.00 · checked 2026-10-15
$240.00a month
- 26
GPT-5.6 Sol · OpenAI
input $120.00 · output $200.00 · checked 2026-10-15
Note: ≤272K prompt tokens; above 272K $8/$30; promotional price through at least Nov 21, 2026
$320.00a month
- 27
Claude Opus 5.5 · Anthropic
input $120.00 · output $200.00 · checked 2026-10-15
$320.00a month
- 28
Claude Opus 4.8 · Anthropic
input $150.00 · output $250.00 · checked 2026-10-15
$400.00a month
- 29
Claude Opus 5 · Anthropic
input $150.00 · output $250.00 · checked 2026-10-15
$400.00a month
- 30
Claude Opus 4.7 · Anthropic
input $150.00 · output $250.00 · checked 2026-10-15
$400.00a month
- 31
Claude Opus 4.6 · Anthropic
input $150.00 · output $250.00 · checked 2026-10-15
$400.00a month
- 32
Claude Opus 4.5 · Anthropic
input $150.00 · output $250.00 · checked 2026-10-15
$400.00a month
- 33
GPT-5.5 · OpenAI
input $150.00 · output $300.00 · checked 2026-10-15
Note: ≤272K prompt tokens; above 272K $10/$45
$450.00a month
- 34
GPT-6 Astra · OpenAI
input $300.00 · output $500.00 · checked 2026-10-15
Note: ≤272K prompt tokens; above 272K $20/$75
$800.00a month
- 35
Claude Fable 5.1 · Anthropic
input $300.00 · output $500.00 · checked 2026-10-15
$800.00a month
- 36
Claude Fable 5 · Anthropic
input $300.00 · output $500.00 · checked 2026-10-15
$800.00a month
- 37
GPT-5.5 Pro · OpenAI
input $900.00 · output $1,800 · checked 2026-10-15
Note: ≤272K prompt tokens; above 272K $60/$270
$2,700a month
Prices: each model provider's own pricing page. Third-party terms apply. Prices checked 2026-10-15; the full table is on /inference. Each bill shows the date its price was last checked against the provider’s page; the table does not record when a price took effect. The bill covers tokens only: it leaves out charges for writing or storing a cache, priority and regional pricing, tool use and taxes.
GPU rental
Is this quote fair?
Place an hourly quote among the current quotes for the same chip and billing model, compared under the same rules as the chip pages. These are posted list prices, not transactions.
Enter a price to place it among 12 comparable quotes for H100: cheapest Voltage Park at $1.99, median $4.245, highest Azure (ND v5) at $12.29.
On-demand list quotes for H100, observed 2026-10-11 to 2026-10-14. These are posted list prices, not transactions. The percentile counts the share of quotes below yours, with equal prices as half.
All 12 comparable quotes
- Voltage Park$1.99 · 2026-10-11
- Hyperstack$3.20 · 2026-10-14
- Crusoe$3.90 · 2026-10-14
- RunPod$3.99 · 2026-10-11
- Lambda$3.99 · 2026-10-14
- Together AI$3.99 · 2026-10-14
- Nebius$4.50 · 2026-10-14
- CoreWeave$6.155 · 2026-10-14
- AWS (p5)$6.88 · 2026-10-14
- Oracle OCI$10.00 · 2026-10-14
- Google Cloud (A3)$11.0613 · 2026-10-14
- Azure (ND v5)$12.29 · 2026-10-14
Offered where at least 5 comparable current quotes exist: B300, B200, H200, H100, A100 80GB and L40S (on-demand list). The on-demand set holds provider-published on-demand list prices from hyperscalers and specialist clouds, compared as the chip pages compare them; capacity blocks, serverless / per-second, marketplace / variable availability and secondary market snapshot quotes are never pooled with them. Each quote keeps its own observation date, listed beside it. Prices: Amazon Web Services; Microsoft Azure Retail Prices API; Google Cloud; Oracle Cloud Infrastructure; each provider's own pricing page; RunPod; Voltage Park. Third-party terms apply.
Commitments
Reserved (1-year / 3-year) list prices
What Azure listed for 1- and 3-year reservations of the ND H100 and H200 VMs whose on-demand prices this site tracks, read from its Retail Prices API on 2026-10-14, beside its pay-as-you-go price from the same read. Azure prices a reservation for the whole term; the hourly figures spread it evenly over 8,760 hours a year. Reservations are their own billing model: they stay out of the on-demand quote table, its medians and the forward curve.
H100 · Reserved (1-year / 3-year) list prices
Azure ND96isr H100 v5, 8× H100 per VM, Linux, US East (same prices in US East 2, US West 2, US West 3).
| Term | Per GPU-hour | vs pay-as-you-go | Azure list price, per VM |
|---|---|---|---|
| Pay-as-you-go | $12.29 | — | $98.32 an hour · since 2023-12-01 |
| 1-year reservation | $7.87 | −36% | $551,221 for the term (≈ $62.92 an hour) · since 2023-12-01 |
| 3-year reservation | $5.40 | −56% | $1,134,310 for the term (≈ $43.16 an hour) · since 2023-12-01 |
On 2026-10-14, Azure listed a 1-year reservation of the ND96isr H100 v5 in US East at the equivalent of $7.87 per GPU-hour, 36% below its $12.29 pay-as-you-go rate, and a 3-year reservation at $5.40, 56% below.
Prices: Microsoft Azure Retail Prices API (pay-as-you-go query, reservation query), captured 2026-10-14 19:19 UTC. Third-party terms apply.
H200 · Reserved (1-year / 3-year) list prices
Azure ND96isr H200 v5, 8× H200 per VM, Linux, US East 2 (same prices in US West 2, US West 3).
| Term | Per GPU-hour | vs pay-as-you-go | Azure list price, per VM |
|---|---|---|---|
| Pay-as-you-go | $10.60 | — | $84.80 an hour · since 2025-04-01 |
| 1-year reservation | $5.82 | −45% | $407,526 for the term (≈ $46.52 an hour) · since 2026-08-01 |
| 3-year reservation | $5.28 | −50% | $1,109,592 for the term (≈ $42.22 an hour) · since 2026-08-01 |
On 2026-10-14, Azure listed a 1-year reservation of the ND96isr H200 v5 in US East 2 at the equivalent of $5.82 per GPU-hour, 45% below its $10.60 pay-as-you-go rate, and a 3-year reservation at $5.28, 50% below.
Prices: Microsoft Azure Retail Prices API (pay-as-you-go query, reservation query), captured 2026-10-14 19:19 UTC. Third-party terms apply.