rag
RAG grounded answer — per job.
About 12,000 input tokens.
No measured data yet Price collection is running, but the benchmark has not executed yet. We do not fill rankings with estimates.
Estimated cost from list prices assumed
Calculated from the collected official price lists — not measured by running the task. Once measurements exist, the measured ranking appears above this table.
Input tokens12,000 Output ratio4% Calls per job1 Cache hit60% Monthly volume30,000
| Model | Input | Output | Per job | Per 1,000 | Monthly |
|---|---|---|---|---|---|
| GLM OCR · zhipu | $0.03 | $0.03 | $0.000374 | $0.37 | $11.23 |
| Gemini 2.5 Flash Lite · google | $0.05 | $0.2 | $0.000408 | $0.41 | $12.24 |
| GLM 4.6v Flashx · zhipu | $0.04 | $0.4 | $0.000413 | $0.41 | $12.38 |
| GLM 4.7 Flashx · zhipu | $0.07 | $0.4 | $0.0006 | $0.60 | $18.00 |
| Gemma 4 E2b · aws_bedrock | $0.05 | $0.1 | $0.000648 | $0.65 | $19.44 |
| Gemini 2.5 Flash Lite Preview · google | $0.1 | $0.4 | $0.000744 | $0.74 | $22.32 |
| Voxtral Mini 1.0 · aws_bedrock | $0.06 | $0.06 | $0.000749 | $0.75 | $22.46 |
| Gemma 3 4B · aws_bedrock | $0.06 | $0.12 | $0.000778 | $0.78 | $23.33 |
| Gemini 2.0 Flash · google | $0.1 | $0.4 | $0.000852 | $0.85 | $25.56 |
| GPT OSS Safeguard 20B · aws_bedrock | $0.072 | $0.206 | $0.000964 | $0.96 | $28.92 |
| Gemini 2.0 Flash Lite · google | $0.075 | $0.3 | $0.001044 | $1.04 | $31.32 |
| Gemini 3.1 Flash Lite · google | $0.125 | $0.75 | $0.00105 | $1.05 | $31.50 |
Why this can differ from your actual bill
- Reasoning tokens are billed as output, but price lists don't say how many each model spends.
- Output length depends heavily on the prompt. Here it is assumed as a fixed ratio of input.
- Cache hit rate depends on your traffic. Applied only to models that publish a cache-read price.
- Batch tiers, long-context tiers and committed-use discounts are not applied.
Price it at your scaleUsers and volume are enough for a monthly figure
Cost estimator →