Tencent Named Its Rivals by Number: The Growth Play Behind Putting GLM-5.3 and Kimi K3 in Your Own Benchmark

Most launches say "best-in-class" and move on.

Tencent's Hy4 preview did something more specific: it named two competitors by name and published exact numbers against them.

What they actually said

Tencent ran its own evaluation — "163 experts and 203 engineering tasks" — and reported Hy4 preview scored "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)."

That's three named models, three exact scores, out of a defined denominator. Not "outperforms leading models." Not a percentage lift with no baseline. A number next to a name next to a number.

They paired it with a second lever: free access to try it yourself, "for two weeks," directly inside WorkBuddy and CodeBuddy — the same products where the benchmark tasks were run.

Why this works

A vague superiority claim asks for trust. A named, numbered comparison asks for verification — and verification is what technical buyers actually want to do before they switch anything.

Naming GLM-5.3 and Kimi K3 specifically also does something a generic "vs. the market" claim can't: it puts Hy4 in the same sentence as models the buyer already has an opinion about, and lets that borrowed context do the positioning work for free.

The two-week trial closes the loop. A number a buyer can't check is a marketing claim. A number a buyer can check for free, this week, inside a tool they already have open, is a fact they'll repeat to their own team.

How to run this yourself

1. Pick two or three competitors your buyer already knows by name — not the whole market, just the names that come up in their own evaluation conversations

2. Run a real internal eval with a defined task count and scoring scale, and publish the denominator, not just the headline number

3. Report their exact scores next to yours, even when the gap is small — a 2.99 vs. 2.94 is more credible than an unsourced "significantly better"

4. Give buyers a free, low-friction way to rerun the comparison themselves, inside the same product you're evaluating

5. Time-box the free access so it drives a decision instead of sitting unused

Bottom line

The growth lesson isn't "publish benchmarks." It's publish benchmarks a specific, named buyer can check themselves, against competitors they already have an opinion about, with a free way to verify it before they trust you.

Sources:

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

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