The fine print, surfaced
Several frontier models charge more per token once your prompt crosses a threshold — GPT-5.6 family doubles input price above 272K tokens, Grok 4.6 jumps at 200K. Comparison sites show the headline number; this page shows the ladder.
Verified 2026-10-21 · source: OpenRouter endpoint pricing overrides
⚠️ Why this matters
A 300K-token codebase analysis on GPT-5.6 Terra Pro costs 2× the headline rate — $4/1M input instead of $2. On a 10-request workload that's the difference between $6 and $12. If your workload lives past the threshold, sort by the tier price, not the base price.
10 models with tiered pricing
| Prompt size | Input /1M | Output /1M | vs base input |
|---|
| Base (below first threshold) | $0.100 | $0.500 | baseline |
| ≥ 100,000 tokens | $0.500 | $2.50 | 5.0× |
| Prompt size | Input /1M | Output /1M | vs base input |
|---|
| Base (below first threshold) | $0.200 | $1.20 | baseline |
| ≥ 272,000 tokens | $0.200 | $0.900 | 1.0× |
| Prompt size | Input /1M | Output /1M | vs base input |
|---|
| Base (below first threshold) | $4.00 | $20.00 | baseline |
| ≥ 272,000 tokens | $4.00 | $15.00 | 1.0× |
| Prompt size | Input /1M | Output /1M | vs base input |
|---|
| Base (below first threshold) | $10.00 | $50.00 | baseline |
| ≥ 272,000 tokens | $10.00 | $37.50 | 1.0× |
| Prompt size | Input /1M | Output /1M | vs base input |
|---|
| Base (below first threshold) | $0.100 | $0.500 | baseline |
| ≥ 272,000 tokens | $0.100 | $0.375 | 1.0× |
| Prompt size | Input /1M | Output /1M | vs base input |
|---|
| Base (below first threshold) | $0.100 | $0.500 | baseline |
| ≥ 272,000 tokens | $0.100 | $0.375 | 1.0× |
| Prompt size | Input /1M | Output /1M | vs base input |
|---|
| Base (below first threshold) | $2.00 | $10.00 | baseline |
| ≥ 272,000 tokens | $2.00 | $7.50 | 1.0× |
| Prompt size | Input /1M | Output /1M | vs base input |
|---|
| Base (below first threshold) | $2.00 | $10.00 | baseline |
| ≥ 272,000 tokens | $2.00 | $7.50 | 1.0× |
| Prompt size | Input /1M | Output /1M | vs base input |
|---|
| Base (below first threshold) | $0.030 | $0.130 | baseline |
| ≥ 32,000 tokens | $0.100 | $0.400 | 3.3× |
| ≥ 256,000 tokens | $0.200 | $0.800 | 6.7× |
| Prompt size | Input /1M | Output /1M | vs base input |
|---|
| Base (below first threshold) | $2.00 | $6.00 | baseline |
| ≥ 200,000 tokens | $4.00 | $12.00 | 2.0× |
Planning a long-context workload?
Models with 1M+ context and flat pricing (Gemini Flash family, Llama 4) often beat tiered frontier models once you're past 272K tokens. Some tiers raise input cost while others reduce output cost, so compare the complete tier rather than assuming every override is a surcharge. Run your numbers in the calculator with your real token counts.