Spend less or ship more with the models you already use.
Move work to a cheaper model without leaving the endpoint you already call.
Every figure here is measured on real Kyma sessions, not on a benchmark.
Calculate savings
Enter what you spend today to see what the move is worth.
$297.62
saved per month
($3,571 per year)
- Spend reduction
- 29.8%
- New monthly spend
- $702.38was $1,000
- New cost per session
- $0.018was $0.025
Modeled on your current model's real sessions, repriced at today's Kyma rates. Cost view keeps the sessions and changes the price; Sessions view keeps your budget. Assumes the same share of input is served from cache on both models.
Measured on real sessions
The calculator reads the same table.
| # | Model | Uptime 30d | Cost / session (p50) | p95 | Tokens / session | Steps / session | tok/s | Sessions |
|---|---|---|---|---|---|---|---|---|
| 1 |  Qwen 3.8 27B | 100.0% | $0.096 | $0.096 | 395,042 | 3 | 49 | 1 |
| 2 | 100.0% | $0.316 | $0.950 | 233,943 | 8 | 17 | 5 | |
| 3 |  Gemma 4 31B | 99.9% | $0.0017 | $0.030 | 21,798 | 2 | 9 | 168 |
| 4 |  DeepSeek V4 Flash | 99.9% | $0.0096 | $0.103 | 125,290 | 9 | 28 | 787 |
| 5 |  MiniMax M2.5 | 99.8% | $0.078 | $0.358 | 234,286 | 8 | 70 | 4 |
| 6 | DeepSeek R1 | 99.5% | $0.0028 | $0.129 | 1,540 | 10 | 73 | 77 |
| 7 | Qwen 3 Coder | 99.5% | $0.018 | $0.490 | 61,006 | 3 | 48 | 68 |
| 8 |  GLM 5.1 | 99.3% | $0.878 | $1.67 | 459,880 | 21 | 29 | 2 |
| 9 |  Gemini 3 Flash | 99.1% | $0.0092 | $0.098 | 11,098 | 5 | 40 | 206 |
| 10 |  Kimi K2.7 Code | 99.1% | $0.054 | $0.389 | 90,050 | 6 | 47 | 9 |
| 11 | Gemini 3.6 Flash | 98.4% | $0.211 | $5.07 | 135,377 | 13 | 29 | 19 |
| 12 | Kimi K2.6 | 98.3% | $0.040 | $0.412 | 61,926 | 4 | 38 | 14 |
| 13 | DeepSeek V4 Pro | 98.2% | $0.021 | $0.273 | 43,260 | 4 | 32 | 174 |
| 14 | Gemini 3.5 Flash Lite | 97.9% | $0.055 | $0.229 | 28,885 | 5 | 54 | 19 |
| 15 | GLM 5.2 | 97.9% | $0.080 | $0.564 | 89,072 | 4 | 83 | 57 |
| 16 |  Sonar | 97.8% | $0.015 | $0.070 | 1,133 | 2 | 26 | 3 |
| 17 |  GPT-5.6 Sol | 97.7% | $0.150 | $0.957 | 48,812 | 21 | 28 | 12 |
| 18 | MiniMax M3 | 97.4% | $0.0077 | $0.014 | 6,659 | 4 | 34 | 2 |
| 19 | Qwen 3.8 Max | 97.3% | $0.0003 | $0.0003 | 135 | 2 | 28 | 1 |
| 20 |  Claude Sonnet 5 | 97.2% | $0.048 | $0.443 | 13,131 | 8 | 18 | 11 |
| 21 | Gemini 3.5 Flash | 97.2% | $0.068 | $0.611 | 15,775 | 4 | 123 | 60 |
| 22 | Claude Fable 5 | 97.0% | $0.368 | $0.956 | 12,914 | 16 | 15 | 4 |
| 23 | GLM 4.5 Air | 96.9% | $0.029 | $0.029 | 150,875 | 5 | 37 | 1 |
| 24 | Gemini 3.7 Flash | 96.8% | $0.119 | $0.398 | 40,413 | 4 | 92 | 12 |
| 25 | Claude Sonnet 4.6 | 96.5% | $0.0067 | $0.0098 | 505 | 6 | 31 | 2 |
| 26 | Gemini 2.5 Flash | 96.4% | $0.016 | $0.087 | 9,137 | 3 | 42 | 231 |
| 27 |  Grok Build | 96.2% | $0.0039 | $0.173 | 1,692 | 2 | 105 | 5 |
| 28 | Grok 4.3 | 96.2% | $0.061 | $0.625 | 110,325 | 3 | 86 | 11 |
| 29 | GLM 4.7 Flash | 96.0% | $0.019 | $0.109 | 78,844 | 5 | 61 | 153 |
| 30 | GPT-5.6 Terra | 95.8% | $0.442 | $0.606 | 154,789 | 3 | 37 | 3 |
| 31 | Claude Haiku 4.5 | 95.6% | $0.0014 | $0.039 | 750 | 4 | 39 | 12 |
| 32 | GPT-5.6 Sol Pro | 95.5% | $0.431 | $0.435 | 141,372 | 2 | 53 | 3 |
| 33 | Claude Opus 5 Fast | 95.4% | $0.030 | $0.370 | 1,805 | 5 | 33 | 9 |
| 34 | Sonar Pro | 95.3% | $0.258 | $0.442 | 2,591 | 32 | 25 | 3 |
| 35 | Gemini 3.1 Pro | 94.9% | $0.101 | $0.101 | 7,855 | 4 | 96 | 1 |
| 36 | Claude Opus 4.7 | 94.7% | $0.212 | $0.269 | 29,383 | 5 | 24 | 3 |
| 37 | GPT-5.6 Luna | 94.4% | $0.0034 | $0.227 | 6,078 | 3 | 39 | 80 |
| 38 |  Muse Spark 1.1 | 94.2% | $0.0002 | $0.0002 | 57 | 2 | 179 | 1 |
| 39 | Llama 4 Maverick | 94.1% | $0.037 | $0.102 | 81,887 | 22 | 33 | 4 |
| 40 | GLM 5.3 Flash | 93.4% | $0.0088 | $0.021 | 107,552 | 2 | 30 | 22 |
| 41 | Kimi K2.5 | 92.7% | $0.145 | $0.273 | 560,204 | 13 | 35 | 2 |
| 42 | GPT-OSS 120B | 92.5% | $0.0014 | $0.033 | 23,648 | 5 | 35 | 102 |
| 43 | GPT-5.6 Luna Pro | 92.4% | $0.044 | $0.253 | 351,449 | 3 | 77 | 10 |
| 44 | MiniMax M2.7 | 91.7% | $0.0004 | $0.050 | 366 | 2 | 54 | 3 |
| 45 | Qwen 3 32B | 91.5% | $0.0015 | $0.0057 | 5,241 | 3 | 45 | 137 |
| 46 | Llama 3.3 70B | 89.9% | $0.0019 | $0.020 | 8,620 | 3 | 16 | 470 |
| 47 | Claude Opus 5 | 89.3% | $0.108 | $5.91 | 19,947 | 4 | 20 | 8 |
| 48 | Qwen 3.7 Flash | 87.5% | $0.014 | $0.047 | 78,574 | 16 | 109 | 130 |
| 49 | GPT-5.6 Terra Pro | 86.7% | $0.060 | $0.448 | 9,877 | 4 | 76 | 6 |
| 50 | Kimi K3 | 85.0% | $0.570 | $0.714 | 490,922 | 18 | 28 | 2 |
| 51 | DeepSeek V3 | 83.0% | $0.011 | $0.040 | 57,337 | 3 | 11 | 160 |
| 52 | Qwen 3.7 Max | 78.3% | $0.048 | $0.158 | 26,040 | 3 | 57 | 61 |
| 53 | Qwen 3.6 Plus | 77.3% | $0.045 | $0.437 | 24,355 | 4 | 50 | 336 |
| 54 | Qwen 3.7 Plus | 59.5% | $0.024 | $0.324 | 16,564 | 4 | 54 | 11 |
| 55 | Qwen 3.8 Flash | 49.1% | $0.020 | $0.097 | 185,042 | 7 | 44 | 12 |
Measured on real Kyma sessions in the last 90 days. A session is the requests from one user and API key, cut at a 5-minute idle gap. Sessions differ per model, so this is what a session costs on that model, not the same task repriced. Uptime is 30-day, one definition, shared with /models and /status. Sessions computed 2026-09-04; prices as of 2026-09-05.