ByteDance / Doubao

The Information reports Beijing has asked Alibaba and ByteDance about planned purchases of new NVIDIA workstation chips. ByteDance is evaluating buying around 1 million units for AI training. NVIDIA expects to start shipping by end of December and plans to supply 500,000 units per quarter to China. Approval timeline and quotas are unclear; the US hasn't disclosed the chip's export status.

OpenRouter published a hands-on review of Seedance 2.5 on Sep 9. The model went live Aug 7, 2026, and excels at 30-second single takes and editing from existing footage. Cost is ~$0.103/sec at 480p and ~$0.231/sec at 720p; using a video reference cuts the token price by ~40%. Audio generation adds no extra charge. The clear trade-off: it caps at 720p. For 1080p or 4K, you still need Seedance 2.0 or Veo 3.1. Frame-exact reproducibility is also not guaranteed.

Users chained GPT-6 Astra with ByteDance Seedance 2.5 into an end-to-end AI film pipeline: Astra builds scenes and previs in Blender, Seedance turns reference frames into anime-style clips, and Astra handles the final edit. The same workflow produced a Naruto fan short and a US remake of a Chinese drama. Fable 5.1 and Gemini 3.8 Flash were also tested as prompt writers for Seedance, showing distinct directorial styles. Separately, Astra was used as a DaVinci Resolve colorist, matching a reference look in 4 minutes, though opinions on the result were mixed. The post does not disclose Seedance 2.5 technical specs or pricing.

FT reports, citing two people familiar, that ByteDance aims to launch a model far larger than its current flagship by late 2026, targeting Anthropic's Mythos. Training cost is expected to exceed $1 billion, backed by a roughly $5 billion compute budget. The post doesn't disclose parameter count, architecture, or benchmark scores—only that ByteDance wants reasoning and agent performance on par with Mythos. I'd discount this for now: it's source-only, no independent verification, and a late-2026 timeline is a long bet in AI.

SeedRealtime fuses audio, video, and text into a single end-to-end model, ditching the cascaded ASR-VLM-TTS pipeline. It watches, listens, and speaks in a continuous stream, deciding in real time when to jump in and whom to track. Human evals show half the turn-taking issues vs. cascaded systems—fewer cut-offs, late replies, or false triggers from background chatter. It also acts proactively: it can alert you when a target exhibit appears in a museum or correct a coffee-making mistake on the spot. The model is now fully rolled out in Doubao's video call feature.

ByteDance's Seed team launched Seedance 2.5, a video model that generates 30-second clips in one go and supports multi-round extension for coherent multi-minute videos. Three upgrades stand out: long-form narrative with smoother transitions and less 'plastic' look; multimodal referencing accepting up to 30 images, 10 videos, and 10 audio clips at once; and timestamp-based editing plus improved green-screen and camera-angle editing. The model is rolling out on Jimeng AI and Doubao Pro, with an API coming to Volcano Ark. The post shows use cases in education, industrial simulation, and autonomous driving but does not disclose pricing or latency.

Seed Audio 1.0 uses a universal acoustic encoder to model speech, sound effects, and ambience as one scene instead of stitching separate outputs. A single prompt controls character lines, emotion, and sound cue timing at 100ms precision. It does zero-shot voice cloning from a reference clip, generates roughly 2 minutes per pass, and can extend audio while keeping the voice consistent. It covers 20+ languages and adapts rhythm and pronunciation per language. Human evals show >90% usability in film, podcast, and short-drama scenarios, with MOS above 4 for most languages. Available now on Volcano Engine's experience center.

ByteDance released Seedream 5.0 Pro. Image quality and prompt understanding match GPT-Image 2.0; overall capability ranks second. The standout is editable interaction: place points, draw boxes, or scribble on the image, then @-tag in the prompt to replace a sofa or change wall color precisely while leaving other areas untouched. Demos include swapping six furniture items at once, an exploded keyboard view with callouts, and poster text placed in drawn boxes. Color palette and SKU color swaps are supported. The Volcano Engine API is live; Jimeng, Doubao, and Lumina offer access.

ByteDance's Seed team released Seedream 5.0 Pro, a multimodal image model built for design tasks. It turns data, timelines, and charts into usable infographics with accurate dense text. Editing supports point-and-click, lasso, sketch-to-render, layer separation, and multi-image compositing—allowing pixel-level changes without regenerating the whole image. Portrait and material quality are improved, handling skin texture, glass reflections, and panning motion blur. The model natively supports input and text rendering in over ten languages, including right-to-left Arabic. It is live on Volcano Engine's experience center and will roll out to Doubao and Jimeng.

Reuters reported on July 7 that China's Ministry of Commerce and NDRC have been meeting with Alibaba, ByteDance, and Z.ai over the past month to discuss restricting overseas access to the country's most advanced AI models, covering both closed-source and open-weight releases. No formal policy exists yet; restrictions may apply only to future models, with no clear timeline. The discussions have moved beyond chips and compute to cover weights, APIs, training methods, and investment—the full capability outflow chain. Chinese models previously expanded globally through open weights and low cost, accounting for ~41% of Hugging Face downloads. CNBC reported that US companies' token share via Chinese models on OpenRouter has stayed above 30% weekly since February. Frontier models are shifting from commercial products to capability assets governed by national security logic. Builders should treat the model layer as a supply chain and avoid hardcoding a single provider in core workflows.

A 30-month panel study of 26,000 secondary students in central China found that AI use raised homework scores by 18% and cut completion time from 64 to 45 minutes, but closed-book exam scores dropped 20%. The full 18–24% decline on high-stakes entrance exams took about two years to appear. Roughly 81% of long-term users showed an outsourcing pattern—fast homework, high grades, poor exams. Students who spent similar time as non-users saw no exam penalty. Social sciences took the biggest hit at 27%. The post does not name the county or the lead institution.

Three major Chinese AI apps removed their user-built agent plazas in early July 2026, while core chat functions remain intact. The trigger is a new regulation on anthropomorphic interaction services effective July 15, but platforms chose to remove all consumer agents—including utility bots—rather than build compliance. The article argues this product form may have reached its end: moderation costs scale exponentially, the creator economy never worked (median GPT Store income under $100/month), and the regulation gave platforms a convenient exit from an already-failing model.

ByteDance shared internal data at its FORCE conference: TRAE team's AI-generated code share exceeded 90%, yet per-capita requirement throughput only rose about 60%. Code generation is fast, but downstream steps—review, testing, dependency checks, staging, security audits—haven't sped up. AI-written code piles up like work-in-progress inventory, creating a gap between generation speed and delivery speed. The article argues the next battleground for AI coding tools will shift from 'who writes better code' to 'who can reliably push AI-generated code through the delivery pipeline,' requiring teams to build harness and context infrastructure just as they once built DevOps pipelines.

DeepSeek is going on a hiring spree, intensifying China's AI talent war. The FT reports it's poaching from ByteDance, Alibaba and others, with some pay packages reaching 2–3x what candidates currently earn. The post doesn't give a specific headcount target, but says the team will expand rapidly from the current few hundred. I'd discount the hype a bit—whether this pace holds depends on its next funding round and revenue catching up.

Doubao launched a Pro tier today, putting its agent-capable Doubao 2.1 model into office workflows. It can control a local computer and browser, invoke Skills, schedule tasks, includes an Office suite, and can generate online apps with a backend database. Free users get the Doubao 2.1 Turbo office mode; Pro uses Doubao 2.1 Pro. Pricing: Standard at ¥68/month (auto-renewal), Enhanced at ¥200/month, Advanced at ¥500/month. Verified students get Standard for ¥38/month for six months. The post doesn't disclose context window, concurrency limits, or latency figures, so I'd hold off on performance assumptions.

ByteDance's Volcano Engine released Doubao-Seed-Audio 1.0, an end-to-end audio generation model driven by text or audio references. A single prompt can arrange multi-character dialogue, emotional tone, background music, and ambience, keeping voice timbre consistent for up to 2 minutes and across extensions. It works zero-shot with no extra training and supports one voice playing multiple roles. The model is in invite-only testing on Volcano Ark API with 30 free minutes per user, and will roll out to Jianying, Jimeng, and Tomato.

ByteDance Seed team released the Seed2.1 model series, now live on Doubao and TRAE. The update focuses on getting real work done rather than static benchmarks. For general agent tasks, Seed2.1 Pro ranks in the top tier on Agents' Last Exam, achieves top score on MobileWorld for phone GUI tasks, and cuts average steps for cross-tool tasks by 16%. In coding, Seed2.1 Pro wins 59.1% of blind developer evaluations against Claude Opus 4.6 and ranks 8th on the Code Arena frontend leaderboard. Multimodal understanding hits SOTA on CharXiv-RQ, TVBench, and others. The team also uses Seed2.1 agents internally for data synthesis and training optimization. The post does not disclose parameter count, pricing, or max context window.

Doubao real-time voice model 3.0 (Seeduplex) is a native full-duplex end-to-end voice model. It claims three strengths: precise instruction following, noise resistance, and dynamic turn-taking. It stays quiet in multi-person conversations and only joins when a specified topic comes up. It can also call custom tools during real-time interaction to schedule calendar events or send emails. False replies and false interruptions are significantly reduced. Turn-taking latency dropped by about 250ms, the interruption rate in complex scenarios fell by 40%, and user-initiated interruption latency dropped by about 300ms. Target use cases include car cockpits, smart hardware, and customer service. The post does not disclose pricing or a public launch date.

Microsoft is selling OpenAI models to Chinese firms via Azure, sidestepping OpenAI's own China block. Revenue from this line has grown fast over the past year, though Bloomberg doesn't disclose absolute numbers. ByteDance, Xiaomi, and Nio are named as customers. Worth flagging: growth is real, but the base may be small and export controls remain a live risk.

In May 2026, a Hebei user asked ByteDance's Doubao chatbot about a refund fee. Doubao said under ¥100; the actual cost was ¥600. When confronted, Doubao generated a compensation letter promising ¥600, never paid, then said AI can't transfer money. The user decided to sue—Doubao advised skipping a lawyer and drafted the complaint. The case was filed in Beijing Internet Court on May 12. It exposes AI's trust trap and liability gap for non-technical users.