Tencent ships Hy Image3.5 at $0.024 an image — parity claim untested

Tencent's image line has been on a two-month clock since it rebuilt the stack in February. The latest tick is a preview model whose rate card does more work than its benchmark does.

Tencent released Hy Image3.5 preview on Tuesday, pricing a 2K image at 0.15 yuan on its domestic cloud API and $0.024 on the international endpoint — about a quarter of what the model it claims to match charges. The model does text-to-image and image-to-image, takes up to five reference images per call for identity and layout control, and supports multi-turn creation and editing that carries the conversation's accumulated context instead of restarting each request. Output is capped at 2K across multiple aspect ratios, and Tencent bills only for images returned — a filtered or failed request is not charged. Tencent says the model is already reachable through TokenHub for developers and inside Yuanbao, ima, Miora, WorkRally, OnSolo and WorkBuddy, plus the company's media-processing and video-on-demand platforms.

The claim attached to the launch is Tencent's own, and it is specific: an internal GSB blind test — hundreds of professional designers scoring paired outputs without knowing which model produced which — put Hy Image3.5 preview 30% ahead of Hy Image3.0, level with ByteDance's Seedream 5.0 Pro, and slightly ahead of Nano-Banana Pro and Alibaba's Qwen-Image-3.0 Pro. Tencent also says the model had a two-month grayscale run inside Yuanbao before launch, during which image-generation requests in that product rose more than 50%. Both numbers are company telemetry, and neither has been reproduced outside Tencent.

That gap is the story underneath the launch. At the published international rate, Seedream 5.0 Pro costs about $90 per 1,000 images; an analysis by OrcaRouter put Hy Image3.5 preview at roughly a tenth of what GPT Image 2.5 Sunburst costs for the same volume — about $1,200 for 50,000 images against about $10,500. None of that settles quality, because the independent board that would settle it had no Hy Image3.5 entry on launch day. What the Artificial Analysis text-to-image arena carried from Tencent was the prior generation: HunyuanImage 3.0 Instruct at 964 Elo and HunyuanImage 3.0 at 945, ranked 65th and 70th. The models Tencent says it has caught sit much higher — Seedream 5.0 Pro at 1078, Qwen-Image-3.0-Pro at 1088, and the board's leader, GPT Image 2.5 Sunburst (max), at 1197. A 30% gain over a model near 945 does not obviously reach 1078, and it does not reach 1197.

The cadence is the more useful signal. Tencent has shipped a numbered image generation roughly every two months since February, each one arriving as a preview before an official release, and it has now put the current preview into six first-party products at once. Read as a schedule rather than a launch, the announcement says Tencent intends to keep landing image generations at a price the rest of the field has not matched, and is willing to ship before the independent boards catch up. Its language-model sibling followed the same playbook — Tencent open-sources Hunyuan Hy4 preview arrived as a preview too.

There is a free window for anyone who wants to form their own opinion: Miora, WorkRally and OnSolo opened a two-week trial on September 22 that closes October 7.

What to watch: the first third-party Elo for Hy Image3.5 — that number, not the 30%, is what tells you whether Tencent closed the gap or priced around it.

Would a four-fold price gap move your image pipeline before an outside benchmark exists? Tell us in the comments.

Akhetonics says its all-optical CPU reaches a customer in 2026

Light-based computing keeps promising more than it delivers — but one Munich startup has just put a date on its bet, and the interview laying it out is doing the rounds on Hacker News this week. Akhetonics says it will deploy its first commercial machine with a major customer by the end of 2026, with several more planned for 2027. The company, founded by Michael Kissner and Leonardo Del Bino, is building a computer where data enters as light, is switched as light, and circulates through memory

The Week in AI — October 5–11, 2026

Every big claim this week turned out to rest on fine print more interesting than the headline: revenue only the company reporting it can define, safety tests sandboxed while the product keeps the web, and a Pentagon phase-out nobody would confirm until reporters kept asking. The week's top 5 1. OpenAI's revenue was $20 billion below the numbers everyone quoted — and the gap was definitional. The Financial Times reported Thursday that OpenAI's annualized revenue runs roughly $20 billion unde

Drone strike shuts a third Yandex data center, taking YandexGPT offline

Russia's largest tech company is learning what the AI era's infrastructure war looks like from the receiving end — three data centers in four days, and with them much of the cloud layer Russian businesses run on. A Ukrainian drone strike knocked out Yandex's data center in Vladimir early Sunday morning, the third of the company's facilities hit since October 8. The site — reported at roughly 50 MW and designed for about 2,880 server racks — stopped operating completely after the attack, Yandex

Agent teams cost up to 5x more, barely score higher

The multi-agent hype train hit a benchmark this weekend — and the grid and the trucking regulators had quiet weeks of their own. Vals AI put agent teams head-to-head with single agents on its Vibe Code Bench, and the teams cost between 1.8 and 5.1 times more for almost no extra quality. The evals company ran GPT-6 Sol and Claude Opus 5.5 solo and in teams across 50 apps at two reasoning efforts; out of four comparisons, only one was statistically significant — Sol at medium effort, where the t

Read more

Akhetonics says its all-optical CPU reaches a customer in 2026

The Week in AI — October 5–11, 2026

Drone strike shuts a third Yandex data center, taking YandexGPT offline

Agent teams cost up to 5x more, barely score higher