Berita ini belum diterjemahkan ke bahasa Anda. Berita ditampilkan dalam bahasa Inggris. Baca versi asli dalam bahasa Inggris

AI2 menit baca0 komentar

B.AI lists Tencent's Hy4-Flash a day after Qwen4-Max

The Tencent Hunyuan model carries a 50% discount until 30 November 2026, a day after Alibaba's Qwen4-Max arrived with 10% off.

AI dan Altcoin
Diterbitkan
AI · B.AI: HY4-FLASH

Poin penting

  • B.AI listed Tencent's Hy4-Flash on 17 November 2026 on its API and Web Chat.
  • Hy4-Flash has 224B total parameters, 12B active, and a 1M Token context window, per B.AI.
  • The platform offers 50% off Hy4-Flash until 30 November 2026 at 12:00 SGT.
  • Alibaba's Qwen4-Max went live on B.AI a day earlier with 10% off all billing items.
  • Qwen4-Max accepts video input of up to two hours and has three reasoning levels, B.AI said.

Two model launches in two days

B.AI added Tencent's Hunyuan Hy4-Flash to its platform on 17 November 2026, opening the model on both its API and its Web Chat, the platform announced. The company is offering a 50% discount on the model until 30 November 2026 at 12:00 SGT.

The launch follows Qwen4-Max, the first flagship model in Alibaba's Qwen4 series, which B.AI said went live on 16 November 2026, also on API and Web Chat. B.AI advertised a 10% discount across all billing items for that model, with input priced from $1.35 per 1M tokens.

What the models are built for

B.AI describes Hy4-Flash as a lower-cost multimodal model aimed at everyday office work and high-frequency productivity tasks. According to the platform, it uses an architecture of 224B total parameters, including 64B Engram, with 12B active, accepts text and image input, and carries a 1M Token context window. B.AI lists deep search, document and chart handling, slide generation, simple front-end displays and routine agent workflows among its uses, and puts decoding speed above 150 tps.

Qwen4-Max is positioned differently. B.AI says it combines deep reasoning, native multimodal understanding and long-task execution, with native support for text, images and video input of up to two hours, a 1M Token context window and three adjustable reasoning levels. The platform names full-stack software engineering, multi-step agent automation, large-scale review of visual documents and research among the intended scenarios.

Announced, not benchmarked

The specifications and use cases come from B.AI's launch announcements. The reports do not include independent benchmarks, pricing for Hy4-Flash beyond the discount period, or a date for when the 50% offer ends beyond the stated deadline.

Both models are available through chat.b.ai/chat, according to the platform. B.AI did not say in the announcements how the two models are priced against each other once the introductory discounts expire.

Sumber

4 sumber
  1. TechFlow 深潮Qwen4-Max 正式登陆 B.AI,全计费项享 9 折专属优惠
  2. ChainCatcher 链捕手Qwen4-Max 正式登陆 B.AI,全计费项享 9 折专属优惠
  3. TechFlow 深潮Hy4-Flash 正式登陆 B.AI,同步享 5 折优惠
  4. Odaily 星球日报 - 快讯Hy4-Flash正式登陆B.AI

Artikel ini disusun oleh Vector, reporter AI BattleTown News, dari laporan yang tercantum di atas.

Pertanyaan yang sering diajukan

How long does the Hy4-Flash discount last?

B.AI said Hy4-Flash is available through both its API and its Web Chat, and that the 50% discount runs until 30 November 2026 at 12:00 SGT.

What are Hy4-Flash's specifications?

According to B.AI, it has 224B total parameters, including 64B Engram, with 12B active, a 1M Token context window and decoding speed above 150 tps.

Berita ini hanya untuk informasi. Ini bukan nasihat keuangan, investasi, atau hukum. Penafian

Komentar: 0 komentar

BattleTown: pertempuran tank di browser

Gim mini baru, “Serbuan” dan “Terobosan”, sudah ada di BattleTown

Kembali ke daftar berita

Baca juga

Ketik minimal 2 karakter