# Cheapest Qwen3.8 27B API providers, compared (prices checked Nov 7, 2026)

URL: https://ernd.ai/inference-prices/qwen3-8-27b

> ernd partners with [Wafer](https://ernd.ai/partners/wafer). Prices are from OpenRouter's public listings.

Qwen3.8 27B is an open-weight model from Qwen. On OpenRouter, 24 listings from 24 providers sell it, and the blended price runs from $0.342 to $1.14 per million tokens.

| Provider | Input /1M | Output /1M | Blended 3:1 | Cache read /1M | Context | Quantization | Latency | Throughput | With ernd card |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| Chutes | $0.039 | $1.25 | $0.342 | $0.0039 | 262K | FP8 | 2.51 s | 28 tok/s | – |
| NEAR AI | $0.04 | $1.35 | $0.368 | $0.018 | 262K | FP8 | 1.26 s | 65 tok/s | – |
| StreamLake | $0.175 | $1.05 | $0.394 | $0.035 | 1.02M | Not listed | 5.24 s | 54 tok/s | – |
| Darkbloom | $0.035 | $1.50 | $0.401 | $0.0175 | 262K | FP4 | 1.37 s | 52 tok/s | – |
| Prime Intellect | $0.034 | $1.86 | $0.492 | $0.031 | 131K | FP8 | 0.85 s | 21 tok/s | – |
| Ionstream | $0.04 | $1.95 | $0.517 | $0.02 | 262K | FP8 | 0.99 s | 46 tok/s | – |
| DeepInfra | $0.15 | $1.88 | $0.581 | $0.0375 | 262K | BF16 | 0.70 s | 34 tok/s | – |
| Modal | $0.04 | $2.30 | $0.605 | $0.02 | 262K | Not listed | 0.96 s | 113 tok/s | – |
| Tenstorrent | $0.15 | $2.00 | $0.613 | $0.05 | 262K | FP8 | 0.61 s | 81 tok/s | – |
| Wafer, ernd partner | $0.06 | $2.30 | $0.62 | $0.059 | 262K | FP8 | 0.80 s | 83 tok/s | $0.558 after 10% cash back |
| DekaLLM | $0.04 | $2.50 | $0.655 | $0.02 | 262K | Not listed | 0.73 s | 89 tok/s | – |
| AkashML | $0.225 | $1.98 | $0.664 | $0.05 | 262K | FP8 | 0.70 s | 43 tok/s | – |
| Parasail | $0.24 | $2.20 | $0.73 | $0.05 | 262K | FP8 | 1.10 s | 54 tok/s | – |
| Reka AI | $0.15 | $2.50 | $0.738 | $0.05 | 262K | FP8 | 0.78 s | 57 tok/s | – |
| Mancer | $0.175 | $2.50 | $0.756 | $0.175 | 262K | FP8 | 0.98 s | 56 tok/s | – |
| ModelRun [by Modular] | $0.20 | $2.50 | $0.775 | $0.09 | 262K | FP4 | 0.28 s | 131 tok/s | – |
| Phala | $0.20 | $2.50 | $0.775 | $0.05 | 1M | Not listed | 1.20 s | 76 tok/s | – |
| Alibaba Cloud Int. | $0.425 | $2.55 | $0.956 | $0.085 | 1M | Not listed | 2.08 s | 39 tok/s | – |
| CoreWeave | $0.40 | $3.00 | $1.05 | $0.15 | 262K | FP8 | 0.61 s | 71 tok/s | – |
| NovitaAI | $0.42 | $3.00 | $1.06 | $0.085 | 1M | Not listed | 4.77 s | 42 tok/s | – |
| Nebius Token Factory | $0.45 | $3.00 | $1.09 | $0.10 | 1M | FP4 | 1.12 s | 123 tok/s | – |
| Cerebras | $0.99 | $1.49 | $1.11 | $0.99 | 66K | FP16 | 0.29 s | 604 tok/s | – |
| Cloudflare | $0.45 | $3.20 | $1.14 | $0.05 | 262K | Not listed | 0.66 s | 36 tok/s | – |
| Venice | $0.45 | $3.20 | $1.14 | – | 262K | FP8 | 1.15 s | 62 tok/s | – |

Blended is three parts input price to one part output, per million tokens. Latency is time to first token and throughput is output tokens per second, both averaged from OpenRouter's daily figures for the last 7 days. Cheapest and fastest only count listings that were up when checked.

Wafer is an ernd partner: 10% back on Wafer inference with the ernd card. With ernd card is the blended price less 10% cash back on Wafer charges. It comes back on the card bill and doesn't change Wafer's list price. It applies when you pay Wafer directly, not through OpenRouter credits.

## Questions

### What is the cheapest Qwen3.8 27B API provider?

Chutes, at $0.342 per million tokens blended ($0.039 input, $1.25 output), in OpenRouter's listings on November 7, 2026. If your workload is mostly output, StreamLake has the lowest output price at $1.05.

### How much does Qwen3.8 27B cost per million tokens?

Input runs from $0.034 to $0.99 per million tokens across 24 listings, and output from $1.05 to $3.20. The middle listing comes to $0.697 blended.

### Which Qwen3.8 27B provider is fastest?

Cerebras had the most throughput over the last 7 days, about 604 tokens per second. ModelRun [by Modular] was quickest to the first token at 0.28 s.

### How do I pay for Qwen3.8 27B inference?

Buy OpenRouter credits and it sends each request to one of these providers, or open an account with a provider and pay it directly. The ernd card is a credit card for US companies. It earns 4% back on AI vendor spend and 1.5% back on everything else. Wafer charges on the card earn 10% back.

Source: OpenRouter public model and provider listings (https://openrouter.ai/models), checked November 7, 2026.

The ernd card is a credit card for US companies. It earns 4% back on AI vendor spend and 1.5% back on everything else. [Request the card at app.ernd.ai](https://app.ernd.ai/login?next=%2Fapp%2Fcard%2Fapply&utm_source=llms&utm_campaign=inference_prices).