Model Cost Advisor
Get three verifiable model stacks based on your use case, volume, and region.
Common cost scenarios
- Local LLM vs API monthly cost calculator
- DeepSeek vs Qwen cost
- China-side alternatives to Claude API cost
- Kimi vs GLM cost
- GPT vs DeepSeek cost
- Gemini vs Claude cost
- Building a domestic API stack
- Overseas API stack cost
- Estimating long-context cost
- Coding-agent model cost
- Batch vs realtime API pricing calculator
AI API token cost calculator
Enter monthly input/output usage to compare same-currency catalogue conditions, or calculate one selected tier or manual rate pair. Calculation runs in your browser without an account or model call.
Continue to model-stack selection →Batch pricing: synchronous and batch token cost calculator
Split one model’s monthly total into synchronous and Batch portions, apply their input/output rates, and compare with an all-synchronous baseline. Batch usage is included in the total, never added twice. No account or model call.
Batch is asynchronous processing for work that does not need immediate responses; it differs from concurrent calls to a synchronous endpoint. Check model eligibility, timing, quotas and current rates. Text tokens only: excludes cache, media, tool calls, retries, tax and FX; no promise of actual savings.
OpenAI Batch API ↗Claude Message Batches ↗
Official processing guides checked 2026-10-15; verify rates separately.
Model API pricing snapshot and conditions
These checked text-token tiers are a snapshot, not live or final quotes. Review each model’s source, date and conditions before selecting a tier in the calculator. Monthly totals do not establish per-request length or peak/off-peak mix.
View rates and official sources for 14 models
| Model / tier | Currency | Input / 1M | Output / 1M | Conditions | Review date / source |
|---|---|---|---|---|---|
| deepseek-flashBase tier | USD | 0.15 | 0.6 | Off-peak, cache-miss input $0.15 / output $0.60 per million tokens. Peak rates are shown separately; cache-hit input is lower. | Official source |
| deepseek-flashPeak, cache miss | USD | 0.3 | 1.2 | All usage at peak rates: Monday–Friday 01:00–04:00 and 06:00–10:00 UTC, excluding Chinese public holidays. Actual peak/off-peak mix is unknown. | Official source |
| deepseek-v4-proBase tier | USD | 0.66 | 1.98 | Off-peak, cache-miss input $0.66 / output $1.98 per million tokens. Peak rates are shown separately. | Official source |
| deepseek-v4-proPeak, cache miss | USD | 1.32 | 3.96 | All usage at peak rates: Monday–Friday 01:00–04:00 and 06:00–10:00 UTC, excluding Chinese public holidays. Actual peak/off-peak mix is unknown. | Official source |
| qwen3.8-maxBase tier | USD | 2 | 6 | International scope: input $2 / output $6 per million tokens, up to 1M input tokens per request. | Official source |
| qwen-flashBase tier | USD | 0.05 | 0.4 | International scope: input $0.05 / output $0.40 per million tokens for requests with at most 256K input tokens. | Official source |
| qwen-flashLong-input tier | USD | 0.25 | 2 | International scope: >256K and ≤1M input tokens per request; the entire request uses this tier. Monthly totals do not determine per-request tiers. | Official source |
| kimi-k2.6Base tier | USD | 0.95 | 4 | Cache-miss input $0.95 / output $4 per million tokens; excludes cache adjustments and taxes. | Official source |
| kimi-k3Base tier | USD | 3 | 15 | Input $3 / output $15 per million tokens; excludes cache read/write charges and taxes. | Official source |
| GLM-5.2Base tier | USD | 1.4 | 4.4 | Z.AI international API: input $1.4 / output $4.4 per million tokens. | Official source |
| GLM-4.7Base tier | USD | 0.6 | 2.2 | Z.AI international API: input $0.6 / output $2.2 per million tokens. | Official source |
| claude-sonnet-5Base tier | USD | 2 | 10 | Anthropic API Sonnet 5: input $2 / output $10 per million tokens. | Official source |
| claude-haiku-4.5Base tier | USD | 1 | 5 | Anthropic API Haiku 4.5: input $1 / output $5 per million tokens. | Official source |
| gpt-5.6-lunaBase tier | USD | 0.2 | 1.2 | Standard short-context input $0.20 / output $1.20 per million tokens, up to 272K input tokens per request. | Official source |
| gpt-5.6-lunaLong-input tier | USD | 0.4 | 1.8 | Requests with >272K input tokens use 2× input and 1.5× output rates for the full request, not based on monthly totals. | Official source |
| gpt-5.6-terraBase tier | USD | 2 | 12 | Standard short-context input $2 / output $12 per million tokens, up to 272K input tokens per request. | Official source |
| gpt-5.6-terraLong-input tier | USD | 4 | 18 | Requests with >272K input tokens use 2× input and 1.5× output rates for the full request, not based on monthly totals. | Official source |
| gemini-3.8-flashBase tier | USD | 0.75 | 3.75 | Paid Standard input $0.75 / output $3.75 per million tokens through 2026-12-31. Verify rates again for later usage. | Official source |
| gemini-3.1-pro-previewBase tier | USD | 2 | 12 | Paid Standard input $2 / output $12 per million tokens for requests with at most 200K input tokens; output includes billed thinking tokens. | Official source |
| gemini-3.1-pro-previewLong-input tier | USD | 4 | 18 | Paid Standard requests with >200K input tokens; output includes billed thinking tokens. Monthly totals do not determine per-request tiers. | Official source |
Listed text tiers only. Tax, FX, cache, batch, media, tool calls and retries may add charges; model access and task quality require separate verification.
Recent approved public records
Inspect inputs, outputs and evidence from other tasks. Publication review does not verify model quality, cost or commercial outcomes.
No approved public records for this tool yet.
FAQ
How do I estimate mixed synchronous and Batch API pricing?
Enter total monthly input/output tokens, the Batch subset, and four same-model, same-currency synchronous/Batch input/output rates. Batch usage is subtracted from synchronous totals before each portion is priced, avoiding double counting. No automatic discount or request-count-to-token assumption.
Is Batch pricing the same as concurrent synchronous requests?
Batch processes jobs asynchronously without requiring immediate responses. Concurrent client calls to synchronous endpoints do not establish Batch-rate eligibility. Check official model, timing, quota and rate conditions before entering eligible usage.
What does Model Advisor do?
It estimates cost and fit from your monthly token volume and scenario against public pricing tiers. Official vendor prices are authoritative; on-site numbers are estimates only.
Are on-site prices final quotes?
No. Vendor official pricing pages are authoritative. On-site figures are estimates for comparing tiers, not contractual quotes.
Why omit exact dollar or RMB figures on scenario pages?
To avoid stale quotes. Scenario pages explain tier differences and selection logic; estimate with your volume in the advisor and confirm on vendor sites.
Does the estimate promise cost savings?
No. Savings depend on real volume, caching, discounts, and vendor price changes. Results are for selection reference only.
Does the cost calculator require an account or model call?
No. Compare same-currency catalogue conditions at the same monthly usage, or calculate a checked tier or manual rate pair in your browser without model quota. Tier assumptions are independent, not added; monthly volume does not establish context or peak mix. Model-stack generation follows current account and tool quotas.
How do scenario pages relate to the tool?
Scenario pages cover common comparison questions and link back to the advisor so you can estimate with your monthly token volume.