Skip to main content
AI Choice Engine

Tencent

Tencent Hy4

Tencent Hy4 is a Tencent model with 1.05M context, $0.834 in / $2.501 out per million tokens, last verified November 4, 2026. Open weights: not verified. Public-board scores are linked to the publisher; vendor-only figures stay in the watch-out, not on the board.

Tencent Hy4 is the full release of Tencent's flagship Hy4 model, adding image understanding and computer use on top of the preview at the same hosted price.

Recorded profile verification November 4, 2026Individual provider fields may changeAccess: ga — Released November 4, 2026 on Tencent Cloud TokenHub (model id hy4) and in Tencent apps such as WorkBuddy, CodeBuddy and Yuanbao; TokenHub service activation and regional endpoint credentials are required. Tencent says weights and system cards follow "in the coming days"; none were published when checked.

AI model specification details
SpecificationTencent Hy4
ProviderTencent
TierFrontier
Context window1.05M
Max output66K
Input / 1M tokens$0.834
Output / 1M tokens$2.501
WeightsNot verifiedUnverified
ParametersNot verifiedUnverified
Reasoning levelsno_think, high
Modalitiestext, image
API model idhy4
ReleasedNovember 4, 2026

Pricing tiers: Tencent Cloud TokenHub list price is $0.834 input/$2.501 output per MTok with $0.042 cache-hit input, unchanged from Hy4 Preview. TokenHub settles Hy4 at 50% of list until 23:59:59 December 31, 2026 (Beijing time); the list price applies after that. The hosted guide rounds context/input/output to 1M/960K/64K; OpenRouter's record of Tencent's own hy4 endpoint lists 1,048,576 context and 65,536 max output (it also shows the 50%-discounted rate). Tencent's announcement rounds the list price to about $0.90/$2.70.

Best for

  • Low-cost frontier API
  • Long-context agent work
  • Chart and image understanding

Watch out

Weights, licence and parameter count are not yet published; do not assume the preview's Apache 2.0 licence or 770B/49B architecture carries over. TokenHub reasoning_effort defaults to high and accepts high or no_think; the preview's none value is not listed for hy4. The WorkBuddy, financial blind-test, GameDev and hallucination figures are Tencent-reported, not independent benchmark runs. The 50% discount and the November 4-15 free WorkBuddy/CodeBuddy offer are temporary.

History

Tencent Hy4 timeline

Release and status events for this model that passed two-source verification — dated, sourced, and linked.

  • Tencent Hy4ReleaseNovember 4, 2026

    Tencent released the full Hy4 on TokenHub and in its apps, adding image understanding and computer use over Hy4 Preview at the same hosted price; weights are promised "in the coming days".

    Tencent Hy4 launch announcement, dated November 4

Source receipts

Catalog figures for this model were checked against the following sources.

Common questions

Tencent Hy4

Answered from the verified figures on this page rather than general guidance.

How much does Tencent Hy4 cost per million tokens?
Tencent Hy4 lists base rates of $0.834 per million input tokens and $2.501 per million output tokens. Estimate both input and output token volume and any additional charges for your workload. Tencent Cloud TokenHub list price is $0.834 input/$2.501 output per MTok with $0.042 cache-hit input, unchanged from Hy4 Preview. TokenHub settles Hy4 at 50% of list until 23:59:59 December 31, 2026 (Beijing time); the list price applies after that. The hosted guide rounds context/input/output to 1M/960K/64K; OpenRouter's record of Tencent's own hy4 endpoint lists 1,048,576 context and 65,536 max output (it also shows the 50%-discounted rate). Tencent's announcement rounds the list price to about $0.90/$2.70.
What is Tencent Hy4's context window?
Tencent Hy4 accepts about 1.05M tokens of context. That only matters if you routinely send very long documents, large codebases, or multi-turn histories that approach that limit.
What is Tencent Hy4 best for?
Tencent Hy4 is a frontier tier from Tencent. It suits low-cost frontier api, long-context agent work, chart and image understanding. Weights, licence and parameter count are not yet published; do not assume the preview's Apache 2.0 licence or 770B/49B architecture carries over. TokenHub reasoning_effort defaults to high and accepts high or no_think; the preview's none value is not listed for hy4. The WorkBuddy, financial blind-test, GameDev and hallucination figures are Tencent-reported, not independent benchmark runs. The 50% discount and the November 4-15 free WorkBuddy/CodeBuddy offer are temporary.
Model comparison catalogBenchmark library