Volume, reliability, latency, cache behaviour and model facts. Aggregated from our own probe and live traffic. No customer identifiers and no per-user numbers. Times are UTC.
Generated 2026-11-04T11:24:12.633544+00:00. Open the JSON.
81,526
Last 30 days
2,819,941,627
Last 30 days
Collecting
First results after 24h of probes
Aggregate usage for the last 30 days. Internal probe traffic is excluded from these totals.
Measured from our own probe and live traffic. A probe is a small fixed request every five minutes, once with a fresh prefix and once repeating a fixed prefix.
Success rate: Collecting data, first results after 24h of probes.
Median TTFT: Not enough data yet.
p95 TTFT: Not enough data yet.
Success rate: Collecting data, first results after 24h of probes.
Median TTFT: Not enough data yet.
p95 TTFT: Not enough data yet.
| Model | Probes | Success | Cold TTFT | Cached TTFT |
|---|---|---|---|---|
| Collecting data, first results after 24h of probes. | ||||
Repeated prompt prefixes are billed at the cached rate. Cached input can never cost more than fresh input.
| Model | Requests | Cached input tokens | Prompt tokens | Cache hit rate |
|---|---|---|---|---|
| deepseek-v4.1-flash-1017 | 32,708 | 0 | 2,388,931,200 | 0.0% |
| deepseek-v4.1-flash | 5,662 | 0 | 232,467,614 | 0.0% |
| deepseek-v4-1-flash-economy-free | 12 | 0 | 1,360,428 | 0.0% |
| qwen3.8-27b | 756 | 0 | 770,126 | 0.0% |
| nvidia-nemotron-3.5-super-vl | 63 | 0 | 24,780 | 0.0% |
| glm-5.3-flash | 174 | 0 | 327,338 | 0.0% |
| deepseek-v4-flash-0731 | 192 | 0 | 250,345 | 0.0% |
| deepseek-v4-pro-ark | 23 | 0 | 296,012 | 0.0% |
| mimo-v2.6-pro | 40 | 0 | 252,918 | 0.0% |
| gemini-4-argon | 20 | 0 | 124,168 | 0.0% |
| kimi-k3 | 43 | 0 | 18,326 | 0.0% |
| step-5-preview | 39 | 0 | 10,006 | 0.0% |
| fikra-embed | 644 | 0 | 23,870 | 0.0% |
| glm-5.3 | 44 | 0 | 2,127 | 0.0% |
| fikra-free-deepseek | 137 | 0 | 18,176 | 0.0% |
| glm-5.2 | 12 | 0 | 458 | 0.0% |
| deepseek-v4-flash-ark | 3 | 0 | 155 | 0.0% |
| deepseek-v4.1-flash-ark | 4 | 0 | 178 | 0.0% |
| fikra-tts | 17 | 0 | 0 | - |
| fikra-stt | 8 | 0 | 0 | - |
| fikra-image | 1 | 0 | 0 | - |
| minimax-h3 | 1 | 0 | 0 | - |
Published only where the value is verified. "Not yet published" is an honest state, not a placeholder.
| Model | Context | Precision |
|---|---|---|
| DeepSeek V4.1 Flash 1017 deepseek-v4.1-flash-1017 | 1,048,576 | FP8 weights, FP4 experts, BF16 activations |
| Deepseek V4.1 Flash deepseek-v4.1-flash | 1,048,576 | FP8 weights, FP4 experts, BF16 activations |
| DeepSeek V4.1 Flash Economy free window deepseek-v4-1-flash-economy-free | 1,048,576 | Not published for this route |
| Qwen 3.8 27B qwen3.8-27b | 1,001,536 | Not published by provider |
| NVIDIA Nemotron 3.5 Super VL nvidia-nemotron-3.5-super-vl | 1,048,576 | BF16 or NVFP4, route not published |
| GLM 5.3 Flash glm-5.3-flash | 327,680 | FP8 (E4M3) |
| Deepseek V4.1 Flash 0731 deepseek-v4-flash-0731 | 1,048,576 | FP8 (E4M3), BF16 activations |
| DeepSeek V4 Pro deepseek-v4-pro-ark | 1,048,576 | FP8 mixed, BF16 activations |
| MiMo V2.6 Pro mimo-v2.6-pro | 1,048,576 | FP8 (E4M3), BF16 activations |
| Gemini 4 Argon gemini-4-argon | 1,048,576 | Not published by provider |
| Kimi K3 kimi-k3 | 1,113,088 | BF16 |
| Step 5 Preview step-5-preview | 1,048,576 | FP8 (E4M3) |
| Fikra Embeddings fikra-embed | - | Not applicable (managed embedding endpoint) |
| GLM 5.3 glm-5.3 | 327,680 | FP8 (E4M3), BF16 activations |
| DeepSeek V4 Flash Free fikra-free-deepseek | 1,048,576 | FP8 (E4M3), BF16 activations |
| GLM 5.2 glm-5.2 | 262,144 | BF16 |
| DeepSeek V4 Flash deepseek-v4-flash-ark | 1,048,576 | FP8 (E4M3), BF16 activations |
| DeepSeek V4.1 Flash deepseek-v4.1-flash-ark | 1,048,576 | FP8 weights, FP4 experts, BF16 activations |
| Fikra Text-to-Speech fikra-tts | - | Not applicable (managed speech endpoint) |
| Fikra Speech-to-Text fikra-stt | - | Not applicable (managed speech endpoint) |
| Fikra Image fikra-image | - | Not applicable (managed image endpoint) |
| MiniMax H3 Video minimax-h3 | - | BF16 (video model) |
Volume is aggregated from stored request usage over the last 30 days, excluding internal probe traffic. Reliability and latency come from a synthetic probe that runs every five minutes and from live request outcomes. Times are UTC. Empty values mean not enough data yet.
Internal probe traffic and load tests are excluded from the public volume figures. Empty values mean not enough data yet.
/api/public/stats returns the same data as JSON, cached for 60 seconds.
Prices: /pricing. Free model: /free-deepseek-api. Setup guides: /guides.