Balanced — strong quality at a mid priceOpen Weights

Hy3 preview Pricing

Tencent · tencent-hy3-preview

ShareX Facebook LinkedIn

Input / 1M tokens

$0.063

Output / 1M tokens

$0.210

Context window

262,144 tokens

≈ 350 pages of text

Max output

—

What it costs in practice

Typical request (1,200 in + 400 out tokens)$0.0002

1,000 requests / month$0.1596/mo 10,000 requests / month$1.60/mo 100,000 requests / month$15.96/mo

Estimate your own workloadOpens the calculator with Hy3 preview preloaded — adjust volume and token counts there.

Price history

List price per 1M tokens since we started tracking in Jun 2026. Each step marks a price change — hover a point for the exact date and rate.

What it excels at

Supports 256k-token context and output at low per-token cost; open weights enable local or private deployment for long-document workloads.

The business tradeoff

Preview status implies potential instability and minimal tooling ecosystem; output tokens carry 4x the input price with unknown real-world latency on varied hardware.

Same tier from other providers

GPT-5.4 MiniOpenAI$0.0027/ typical request Claude Sonnet 4.6Anthropic$0.0096/ typical request Llama 4 MaverickMeta$0.0006/ typical request

Head-to-head

Hy3 preview vs GPT-5.4 Mini →Hy3 preview vs Claude Sonnet 4.6 →Hy3 preview vs Llama 4 Maverick →

Vendor list rates, as of Jul 23, 2026 · source: openrouter · per-request examples assume 1,200 input + 400 output tokens.