Together AI
GPT OSS 20B (Together) token cost
One of the cheapest open GPT-class hosts — GPT OSS 20B on Together.
Published list rates
Input
$0.05
per 1M tokens
Output
$0.2
per 1M tokens
Context window
131,072
tokens
Pricing source: Together AI pricing
Estimate cost
Estimate from text
Calculate from a token total
Project monthly volume
Daily requests × tokens per request × 30 days.
Estimated cost
- Estimated tokens
- —
- Characters
- —
- Words
- —
- Input tokens
- —
- Output tokens
- —
- Input cost
- —
- Output cost
- —
- Total cost
- —
- Combined price per 1M tokens
- —
- Estimated monthly cost
- —
- Estimated daily cost
- —
Cost as volume grows
What GPT OSS 20B (Together) costs if you only send tokens, only get a reply, or do both (70% in / 30% out). Rows start small and go up to huge traffic.
| Tokens | Input only | Output only | Mix (70% in / 30% out) |
|---|---|---|---|
| 1 | $0.00000005 | $0.0000002 | $0.000000095 |
| 10 | $0.0000005 | $0.000002 | $0.00000095 |
| 100 | $0.000005 | $0.00002 | $0.00001 |
| 200 | $0.00001 | $0.00004 | $0.000019 |
| 300 | $0.000015 | $0.00006 | $0.000029 |
| 400 | $0.00002 | $0.00008 | $0.000038 |
| 500 | $0.000025 | $0.0001 | $0.000048 |
| 1,000 | $0.00005 | $0.0002 | $0.000095 |
| 2,000 | $0.0001 | $0.0004 | $0.00019 |
| 5,000 | $0.00025 | $0.001 | $0.000475 |
| 10,000 | $0.0005 | $0.002 | $0.00095 |
| 25,000 | $0.00125 | $0.005 | $0.002375 |
| 50,000 | $0.0025 | $0.01 | $0.00475 |
| 100,000 | $0.005 | $0.02 | $0.0095 |
| 250,000 | $0.0125 | $0.05 | $0.02375 |
| 500,000 | $0.025 | $0.1 | $0.0475 |
| 1,000,000 | $0.05 | $0.2 | $0.095 |
| 5,000,000 | $0.25 | $1.00 | $0.475 |
| 10,000,000 | $0.5 | $2.00 | $0.95 |
| 50,000,000 | $2.50 | $10.00 | $4.75 |
| 100,000,000 | $5.00 | $20.00 | $9.50 |
| 500,000,000 | $25.00 | $100.00 | $47.50 |
| 1,000,000,000 | $50.00 | $200.00 | $95.00 |
| 10,000,000,000 | $500.00 | $2,000.00 | $950.00 |
| 100,000,000,000 | $5,000.00 | $20,000.00 | $9,500.00 |
Tokens for a fixed budget
How far a fixed spend goes on GPT OSS 20B (Together) at these list rates.
| Budget | Input tokens | Output tokens | Mixed tokens (70/30) |
|---|---|---|---|
| $0.01 | 200,000 | 50,000 | 105,263 |
| $0.1 | 2M | 500,000 | 1.05M |
| $0.5 | 10M | 2.5M | 5.26M |
| $1.00 | 20M | 5M | 10.5M |
| $5.00 | 100M | 25M | 52.6M |
| $10.00 | 200M | 50M | 105.3M |
| $20.00 | 400M | 100M | 210.5M |
| $25.00 | 500M | 125M | 263.2M |
| $30.00 | 600M | 150M | 315.8M |
| $50.00 | 1B | 250M | 526.3M |
| $100.00 | 2B | 500M | 1.05B |
| $250.00 | 5B | 1.25B | 2.63B |
| $500.00 | 10B | 2.5B | 5.26B |
| $1,000.00 | 20B | 5B | 10.5B |
| $5,000.00 | 100B | 25B | 52.6B |
| $10,000.00 | 200B | 50B | 105.3B |
At these rates
- A workload with 70% input and 30% output costs $0.095 per 1 million total tokens and $9.50 per 100 million.
- At the same token count, output costs 4 times the input rate. Response length therefore deserves its own budget limit.
- Filling the 131,072-token context window with input alone would cost about $0.006554, before any output.
These figures use public list prices. Cache eligibility, batch or priority modes, long-context surcharges, discounts, and regional billing can change the invoice.