MODEL SELECTION / API WORKBENCH
AI API Cost Calculator
Know your usage. Choose your model.
One workload, different bills. Scale a single request to a month and compare model cost with task performance.
01 / Set your workload
Pick an illustrative scenario, then adjust your usage.
Missing cache prices use the standard input rate. Cache writes are extra.
- Monthly requests
- 3,000
- Monthly input tokens
- 24,000,000
- Monthly output tokens
- 6,000,000
How many tokens is that?
Rough heuristic: ~1 token per CJK character, or 4 other characters per token. Use the provider’s tokenizer for exact counts.
≈ 0 Token · 0 characters
CURRENT FILTERS · LOWEST ESTIMATE
$5.04
/ monthDeepSeek V4 Flash Preview
Lower cost does not guarantee a better fit. Evaluate with your own tasks.
Same usage × 12 months $60.48
Lowest estimate, broken down
- Uncached input
- $3.36
- Cached input
- $0.00
- Output
- $1.68
02 / Compare models
Current task leaderboard and supplemental models only; not every model on the market.
Exchange rate date: 2026-10-16
35 models · 30 priced · USD / million tokens
- DeepSeek V4 Flash Preview
DeepSeek · Task rank #31 · Score 36.3
$5.04
/ month
Input $0.14Cached $0.0028Output $0.28Price checked 2026-09-26Official pricing ↗Cost breakdown
- Uncached input
- $3.36
- Cached input
- $0.00
- Output
- $1.68
Per request: $0.00168
- GLM-5.3 Flash
Z.ai · Task rank #14 · Score 70.1
$5.38
/ month
Input $0.12Cached $0.034399Output $0.42Price checked 2026-09-26Official pricing ↗Cost breakdown
- Uncached input
- $2.87
- Cached input
- $0.00
- Output
- $2.51
Per request: $0.001795
- GPT-6 Luna
OpenAI · Task rank #25 · Score 43.6
$5.40
/ month
Input $0.10Cached $0.01Output $0.50Price checked 2026-09-26Official pricing ↗Cost breakdown
- Uncached input
- $2.40
- Cached input
- $0.00
- Output
- $3.00
Per request: $0.0018
- GPT-5.6 Luna
OpenAI · Task rank #26 · Score 43.6
$12.00
/ month
Input $0.20Cached $0.02Output $1.20Price checked 2026-09-26Official pricing ↗Cost breakdown
- Uncached input
- $4.80
- Cached input
- $0.00
- Output
- $7.20
Per request: $0.004
- DeepSeek V4.1 Flash
DeepSeek · Task rank #8 · Score 87.1
$14.40
/ month
Input $0.30Cached $0.006Output $1.20Price checked 2026-09-26Official pricing ↗Cost breakdown
- Uncached input
- $7.20
- Cached input
- $0.00
- Output
- $7.20
Per request: $0.0048
- DeepSeek V4 Pro Preview
DeepSeek · Task rank #28 · Score 39.1
$15.66
/ month
Input $0.44Cached $0.003625Output $0.87Price checked 2026-09-26Official pricing ↗Cost breakdown
- Uncached input
- $10.44
- Cached input
- $0.00
- Output
- $5.22
Per request: $0.00522
- DeepSeek V4 Flash 0731
DeepSeek · Task rank #36 · Score 18.1
$18.84
/ month
Input $0.45Cached $0.014956Output $1.35Price checked 2026-09-26Official pricing ↗Cost breakdown
- Uncached input
- $10.77
- Cached input
- $0.00
- Output
- $8.08
Per request: $0.006282
- Gemini 3.8 Flash
Google · Task rank #7 · Score 88.0
$40.50
/ month
Input $0.75Cached $0.075Output $3.75Price checked 2026-09-26Official pricing ↗Cost breakdown
- Uncached input
- $18.00
- Cached input
- $0.00
- Output
- $22.50
Per request: $0.0135
- Gemini 3.7 Flash
Google · Task rank #12 · Score 75.5
$40.50
/ month
Input $0.75Cached $0.075Output $3.75Price checked 2026-09-26Official pricing ↗Cost breakdown
- Uncached input
- $18.00
- Cached input
- $0.00
- Output
- $22.50
Per request: $0.0135
- GLM-5.2
Z.ai · Task rank #33 · Score 34.0
$53.84
/ month
Input $1.20Cached $0.30Output $4.19Price checked 2026-09-26Official pricing ↗Cost breakdown
- Uncached input
- $28.72
- Cached input
- $0.00
- Output
- $25.13
Per request: $0.017947
- Muse Spark 1.1
Meta · Task rank #17 · Score 65.8
$55.50
/ month
Input $1.25Cached —Output $4.25Price checked 2026-09-26Official pricing ↗Cost breakdown
- Uncached input
- $30.00
- Cached input
- $0.00
- Output
- $25.50
Per request: $0.0185
- Muse Spark 1.2
Meta · Task rank #18 · Score 65.6
$55.50
/ month
Input $1.25Cached —Output $4.25Price checked 2026-09-26Official pricing ↗Cost breakdown
- Uncached input
- $30.00
- Cached input
- $0.00
- Output
- $25.50
Per request: $0.0185
Cost vs. performance: the frontier
Cost–performance frontier
Models on the frontier have no alternative with a lower cost and an equal or better score.Adjust your usage to see how the trade-offs change.
On the current frontier. Evaluate it with your actual tasks.
Per million tokens: Input $0.14 · Cache $0.0028 · Output $0.28
Task scores are comparative indices, not accuracy or success rates. The cost axis is logarithmic. Only models with complete pricing are compared.
How is the bill calculated?
Monthly cost = [input tokens × (cache miss rate × input price + cache hit rate × cache price) + output tokens × output price] × daily requests × days ÷ 1,000,000.
Text token costs only. Include billed reasoning tokens in output usage. Excludes tools, search, images, audio, cache writes, tax, subscriptions, tiered prices and batch discounts. Missing prices are never treated as zero. Annual figures multiply the same monthly usage by 12.