Tencentopen weightsHy3 pricing
$0.13 per 1M input tokens and $0.53 per 1M output tokens. In real work that is 0.07¢ per chat answer and 0.28¢ to summarize a 20-page PDF.
What Hy3 costs per task
Token counts estimated from a measured model of the same lab (Claude Haiku 5.5), rescaled by tokenizer. Reasoning tokens are billed as output.
| Task | Cost | Tokens used | Price rank | Basis |
|---|---|---|---|---|
| Answer a chat question | 0.07¢$0.69 / 1,000 | 61 in · 1,055 out · 244 reasoning | 78 of 336 | estimated |
| Reply to a customer email | 0.04¢$0.38 / 1,000 | 362 in · 290 out · 346 reasoning | 90 of 336 | estimated |
| Write a 1,500-word blog post | 0.20¢$2.00 / 1,000 | 90 in · 3,560 out · 200 reasoning | 55 of 336 | estimated |
| Write 10 product descriptions | 0.22¢$2.25 / 1,000 | 454 in · 1,803 out · 2,342 reasoning | 81 of 336 | estimated |
| Translate 1,000 words | 0.14¢$1.38 / 1,000 | 1,832 in · 2,153 out | 81 of 336 | estimated |
| Proofread one page (500 words) | 0.08¢$0.77 / 1,000 | 987 in · 913 out · 304 reasoning | 82 of 336 | estimated |
| Summarize a 20-page PDF | 0.28¢$2.79 / 1,000 | 18,004 in · 672 out · 118 reasoning | 83 of 336 | estimated |
| Extract data from an invoice | 0.07¢$0.69 / 1,000 | 546 in · 574 out · 591 reasoning | 74 of 336 | estimated |
| Analyze a 200-row spreadsheet | ≥0.92¢$9.17 / 1,000 | 5,433 in · 566 out · 15,434 reasoning | 101 of 336 | estimated |
| Answer from 5 retrieved documents (RAG) | 0.17¢$1.66 / 1,000 | 5,184 in · 1,096 out · 753 reasoning | 89 of 336 | estimated |
| Classify 100 support tickets | 0.10¢$1.02 / 1,000 | 1,747 in · 535 out · 961 reasoning | 31 of 336 | estimated |
| Review 500 lines of code | 0.72¢$7.16 / 1,000 | 7,117 in · 3,258 out · 8,532 reasoning | 113 of 336 | estimated |
| Write a SQL query | 0.09¢$0.90 / 1,000 | 182 in · 840 out · 810 reasoning | 91 of 336 | estimated |
| Run a 50-step coding agent | 14¢$140 / 1,000 | 3,101,175 in · 20,790 out · 29,550 reasoning | 74 of 336 | modeled |
Price at every provider
List prices per 1M tokens (input / output) at each host serving Hy3 through OpenRouter today. Hosts can differ in quantization, context and speed.
| Provider | Input / output | Cached input | Context | Uptime |
|---|---|---|---|---|
| T Tencentfp8 | $0.13 / $0.53 | $0.033 | 262K | 100% |
 Novitaunknown | $0.14 / $0.58 | $0.035 | 262K | 100% |
| G GMICloudbf16 | $0.14 / $0.58 | $0.035 | 262K | 100% |
| P Phalaunknown | $0.15 / $0.64 | $0.04 | 262K | |
| A AtlasCloudfp8 | $0.20 / $0.80 | $0.05 | 262K | 100% |
Price history
First seen in our price history on Jul 17, 2026 at $0.20 input / $0.80 output. Since then the output price moved from $0.80 to $0.53 per 1M tokens (5 changes).
FAQ
How much does Hy3 cost?
Hy3 costs $0.13 per 1 million input tokens and $0.53 per 1 million output tokens, with cached input at $0.033 (list price on Oct 18, 2026). In practice one chat answer costs about 0.07¢ and summarizing a 20-page PDF about 0.28¢.
Where is Hy3 cheapest?
Among 5 providers we track, the lowest list price today is at Tencent: $0.13 input and $0.53 output per 1M tokens.
Does Hy3 charge for reasoning tokens?
Yes. Reasoning (thinking) tokens are billed as output tokens. That is why the cost per task can be much higher than the visible answer suggests - our task costs include them.
Is the Hy3 cost per task measured or estimated?
Estimated: Hy3 was not run directly. Tokens come from Claude Haiku 5.5 and are rescaled by the tokenizer of Tencent. Real usage can differ.