Tencent Hy4's 1M-Token Context at Under $1/Million Input Tokens Creates a New Service: Rebuilding Document Workflows That Used to Need Chunking

Tencent just made long documents cheap.

Hy4 preview ships with "770B total parameters and 49B active parameters" and a context window "exceeding 1M tokens" — at a price of "USD 0.834 per million input tokens, USD 2.501 per million output tokens and USD 0.042 per million tokens for cache hits."

That combination — a context window that swallows entire codebases or research libraries, at input pricing under a dollar per million tokens — turns a technical spec into a business opportunity.

What actually shipped

Tencent's own internal evaluation, run across "163 experts and 203 engineering tasks," scored Hy4 preview "an average of 2.99 out of 4.00" — ahead of GLM-5.3 (2.92/4.00) and Kimi K3 (2.94/4.00). The model is available through "WorkBuddy and CodeBuddy, as well as Yuanbao, ima and other Tencent products," with free access on WorkBuddy and CodeBuddy "for two weeks."

Tencent also reports a "31.8% compared with the baseline" throughput increase from autonomous inference optimization, and points to strength in "coding, office work, and scientific research" — specifically long-context development and cross-document collaboration.

The service gap this opens

Most teams still process long documents the slow way: chunking, re-summarizing, losing context between passes, or paying per-seat for a document AI tool that caps out well under 1M tokens.

A 1M+ token window priced under a dollar per million input tokens removes the technical excuse for that. What's left is packaging:

• Law firms and compliance teams that need to reason across an entire contract set, not one PDF at a time
• Engineering teams that want a single pass over a full monorepo instead of RAG'd snippets
• Research groups that need cross-document synthesis across dozens of papers at once

None of these buyers want "a bigger context window." They want the specific workflow solved end to end.

Money Play

1. Pick one document-heavy workflow that currently breaks past a normal context window — full-repo code review, multi-contract due diligence, or literature review synthesis

2. Build a thin pipeline around Hy4's API that ingests the full document set in one pass instead of chunking, and benchmark the output quality against the client's current chunked process

3. Price the pilot as a fixed-scope proof: "we'll process your last quarter's contracts, codebase, or papers in one pass and show you what chunking missed"

4. Use the two-week free WorkBuddy/CodeBuddy access window to build and demo the pilot at near-zero cost before quoting a paid engagement

5. Turn the pilot into a monthly retainer: ingestion pipeline maintenance, model routing (Hy4 for long-context passes, a cheaper model for short tasks), and output QA

Bottom line

Tencent didn't just release another model. It released a price point and a context length that make "just read the whole thing" viable for workflows that used to require expensive workarounds. The business isn't the model — it's being the one who rebuilds the workflow around it before your client's competitor does.

Sources:

https://www.tencent.com/tencent-releases-and-open-sources-tencent-hy4-preview/

https://news.ycombinator.com/item?id=49492632