Pricing
You pay for tokens. The prices are published.
Every model in the catalog has a list price per million tokens for input, cached input, and output. The same numbers appear on the model page, in the rate sheet below, and on the usage the Sylphx console reports — so a bill is never a surprise.
Models listed4612 providers
Cheapest input$0.0462inception/mercury-2.5
Largest context1.05Mtokens per request
MeteringPer tokenon every Responses call
Estimate a monthly bill
Pick a model and a token mix. The estimate uses the same list prices as the rate sheet — no rounding tricks, no hidden platform fee in the number.
2,000,000 tokens
500,000 tokens
Estimated spend
$10.40
Input side
$4.62
Output side
$5.78
Estimate only: it uses today’s list prices for the selected model and assumes every token is billed at the input or output rate. Cached input is cheaper than the input rate, and failed requests that return an error status are not billed as completions.
How usage is metered
- One meter on Responses. Each response reports the input and output tokens it used, and that count is what the price applies to.
- Cached input is discounted. When a request reuses a prompt prefix, the cached tokens are billed at the cached-input rate shown for that model.
- Retries do not double-bill. Send an Idempotency-Key and a retried create replays the original response instead of charging again.
- Read your own numbers. The Sylphx console shows requests, tokens, and settled spend for the workspace, broken down by model.
- These are our own prices. We sell each model at the rate on this page, and the meter bills the same numbers — one rate sheet, no per-model markup table to decode and no fee added at checkout.
Charged against your Sylphx API key at https://api.sylphx.ai/v1. Keys are prefixed sylphx_sk_ and can be revoked at any time in the Sylphx console.
The rate sheet
USD per million tokens, per model. Cached input is what you pay when a prefix is reused; the percentage under it is the saving against the input rate.
46 of 46 models
| Model | Input / M | Cached input / M | Output / M | Context | Est. mix / M |
|---|---|---|---|---|---|
| Claude Fable 5.1anthropic/claude-fable-5.1 | $10.00 | $0.25−98% | $50.00 | 1.0M | ~$20.00 |
| Claude Haiku 5.5anthropic/claude-haiku-5.5new | $0.5775 | $0.05775−90% | $2.888 | 1.0M | ~$1.155 |
| Claude Opus 5.5anthropic/claude-opus-5.5new | $4.62 | $0.231−95% | $23.10 | 1.0M | ~$9.24 |
| Claude Sonnet 5.5anthropic/claude-sonnet-5.5new | $2.31 | $0.1155−95% | $11.55 | 1.0M | ~$4.62 |
| DeepSeek V4 Prodeepseek/deepseek-v4-pro | $1.206 | $0.2412−80% | $2.412 | 1.0M | ~$1.507 |
| DeepSeek V4.1 Flashdeepseek/deepseek-v4.1-flash | $0.3 | $0.02079−93% | $1.20 | 1.05M | ~$0.525 |
| Gemini 3.5 Flash Litegoogle/gemini-3.5-flash-lite | $0.3 | $0.03−90% | $2.50 | 1.05M | ~$0.85 |
| Gemini 3.8 Flashgoogle/gemini-3.8-flash | $0.75 | $0.075−90% | $3.75 | 1.05M | ~$1.50 |
| Gemini 4 Argongoogle/gemini-4-argonnew | $2.31 | $0.1155−95% | $11.55 | 1.05M | ~$4.62 |
| GLM-5.3z-ai/glm-5.3 | $1.40 | $0.26−81% | $4.40 | 1.05M | ~$2.15 |
| GLM-5.3-Flashz-ai/glm-5.3-flash | $0.15 | $0.03465−77% | $0.5 | 1.05M | ~$0.2375 |
| GLM-5.3-FlashXz-ai/glm-5.3-flashxnew | $0.4274 | $0.104−76% | $1.444 | 1.0M | ~$0.6815 |
| GPT Audioopenai/gpt-audio | $2.50 | $0.25−90% | $10.00 | 128K | ~$4.375 |
| GPT Audio Miniopenai/gpt-audio-mini | $0.6 | $0.06−90% | $2.40 | 128K | ~$1.05 |
| GPT-5.4 Image 2openai/gpt-5.4-image-2 | $8.00 | $2.00−75% | $15.00 | 272K | ~$9.75 |
| GPT-5.5openai/gpt-5.5 | $11.00 | $1.10−90% | $49.50 | 1.05M | ~$20.625 |
| GPT-5.5 Proopenai/gpt-5.5-pro | $66.00 | $6.60−90% | $297.00 | 1.05M | ~$123.75 |
| GPT-5.6 Terraopenai/gpt-5.6-terra | $4.40 | $0.44−90% | $19.80 | 1.05M | ~$8.25 |
| GPT-5.6 Terra Proopenai/gpt-5.6-terra-pro | $4.62 | $0.462−90% | $20.79 | 1.05M | ~$8.662 |
| GPT-6 Astraopenai/gpt-6-astra | $22.00 | $2.20−90% | $82.50 | 1.05M | ~$37.125 |
| GPT-6 Astra Proopenai/gpt-6-astra-pro | $23.10 | $2.31−90% | $86.625 | 1.05M | ~$38.981 |
| GPT-6 Lunaopenai/gpt-6-lunanew | $0.22 | $0.022−90% | $0.825 | 1.05M | ~$0.3712 |
| GPT-6 Luna Proopenai/gpt-6-luna-pronew | $0.231 | $0.0231−90% | $0.8662 | 1.05M | ~$0.3898 |
| GPT-6.1 Solopenai/gpt-6.1-solnew | $4.40 | $0.22−95% | $16.50 | 1.05M | ~$7.425 |
| GPT-6.1 Sol Proopenai/gpt-6.1-sol-pronew | $4.62 | $0.231−95% | $17.325 | 1.05M | ~$7.796 |
| Grok 4.7x-ai/grok-4.7new | $4.00 | $1.00−75% | $12.00 | 500K | ~$6.00 |
| Grok Build 0.1x-ai/grok-build-0.1 | $2.31 | $0.462−80% | $4.62 | 256K | ~$2.888 |
| Kimi K2.7 Codemoonshotai/kimi-k2.7-code | $0.95 | $0.2079−78% | $4.00 | 262K | ~$1.712 |
| Kimi K3moonshotai/kimi-k3 | $3.00 | $0.3−90% | $15.00 | 1.05M | ~$6.00 |
| Mercury 2.5inception/mercury-2.5 | $0.0462 | $0.00462−90% | $0.1732 | 260K | ~$0.07796 |
| MiMo-V2.6-Flashxiaomi/mimo-v2.6-flashnew | $0.14 | $0.0028−98% | $0.28 | 1.05M | ~$0.175 |
| MiMo-V2.6-Proxiaomi/mimo-v2.6-pronew | $0.435 | $0.0036−99% | $0.87 | 1.05M | ~$0.5437 |
| MiMo-V2.6-Pro-UltraSpeedxiaomi/mimo-v2.6-pro-ultraspeednew | $4.35 | $0.036−99% | $8.70 | 1.05M | ~$5.438 |
| MiniMax-M3minimax/minimax-m3 | $0.6 | $0.12−80% | $2.40 | 1.0M | ~$1.05 |
| Muse Glimmer 30Bmeta/muse-glimmer-30b | $0.3465 | $0.0462−87% | $1.386 | 131K | ~$0.6064 |
| Muse Spark 1.3meta/muse-spark-1.3 | $1.444 | $0.1732−88% | $4.909 | 1.05M | ~$2.31 |
| Nano Banana 2google/gemini-3.1-flash-image | $0.5775 | $0.05775−90% | $60.00 | 131K | ~$15.433 |
| Nano Banana 2 Litegoogle/gemini-3.1-flash-lite-image | $0.2888 | $0.02888−90% | $30.00 | 66K | ~$7.717 |
| Nano Banana 2.1google/gemini-nano-banana-2.1new | $1.50 | $0.15−90% | $7.50 | 66K | ~$3.00 |
| Nano Banana Progoogle/gemini-3-pro-image | $2.31 | $0.231−90% | $120.00 | 66K | ~$31.733 |
| Qwen3.6 35B A3Bqwen/qwen3.6-35b-a3b | $0.248 | $0.05775−77% | $1.485 | 262K | ~$0.5573 |
| Qwen3.7 Plusqwen/qwen3.7-plus | $1.20 | $0.12−90% | $4.80 | 1.0M | ~$2.10 |
| Qwen3.8 2.4T A95Bqwen/qwen3.8-2.4t-a95b | $2.31 | $0.2888−88% | $6.93 | 1.05M | ~$3.465 |
| Qwen3.8 27Bqwen/qwen3.8-27b | $0.425 | $0.085−80% | $2.55 | 1.0M | ~$0.9562 |
| Qwen3.8 Flashqwen/qwen3.8-flash | $0.165 | $0.0176−89% | $0.517 | 1.0M | ~$0.253 |
| Qwen3.8 Maxqwen/qwen3.8-max | $2.00 | $0.25−88% | $6.00 | 1.0M | ~$3.00 |
“Est. mix / M” is a 3:1 input:output estimate for a million tokens, shown to compare models. Your bill uses the exact input, cached-input, and output counts on each response.
Billing questions
The short answers. For the request-and-response details, start at the quickstart.
What exactly am I charged for?
Tokens processed on Responses calls: input tokens at the input rate, cached prefix tokens at the cached-input rate, and generated tokens at the output rate for the model you named.
Is there a subscription or a minimum?
Pricing on this API is usage-based: you are charged for the tokens you use, at the prices above. There is no per-model access fee listed here, and adding a key costs nothing.
Do failed requests cost anything?
A request that fails validation or returns an error status does not produce a completed response to bill. Retries are safe when you send an Idempotency-Key: a replayed create returns the original response instead of charging a second time.
How do prices change?
The catalog is the rate sheet: a price change publishes on the model page, the pricing page, and the API’s own model payload at the same time. Historic usage keeps the price that applied when it was metered.
Where can I see what I have spent?
The Sylphx console shows settled spend, request counts, and token totals for your workspace, with a per-model breakdown so you can see which model is driving the bill.
Start with one keySend a first request
Lowest listed input price: Mercury 2.5 at $0.0462 per million tokens.