Price: $0.05517 0.1449%
Market Cap: $8.65B 0.416%
Volume (24h): 400.64M 0%
Dominance: 0.416%
Price: $0.05517 0.1449%
Market Cap: $8.65B 0.416%
Volume (24h): 400.64M 0%
Dominance: 0.416% 0.416%
  • Price: $0.05517 0.1449%
  • Market Cap: 8.65B 0.416%
  • Volume (24h): 400.64M 0%
  • Dominance: 0.416% 0.416%
  • Price: $0.05517 0.1449%
Home > 视频 > Hype or Hire: Hy4 vs a 61k-token invoice archive

Hype or Hire: Hy4 vs a 61k-token invoice archive

Release: 2026/11/05 12:00 Reading: 0

Original author:Hype or Hire

Original source:https://www.youtube.com/embed/19fxZHgHKZM

Tencent shipped Hy4 on November 4 with a promised one million token context window. We gave it sixty one thousand tokens of accounts payable paperwork, and on four straight tasks it returned an empty message. A legal-domain model at the same price finished three of the four tasks completely. In this video: - Three models, four real finance tasks, seven scoreable answers - The invoice marked paid whose wire was returned by the bank - The rate change that only applies from July, and the fee that is not in the contract - The renewal deadline that moved from sixty days to ninety days in the last document of the archive - Why the newest model produced no answer at 900, 4,500 or 12,000 output tokens How we tested: - Models, exact OpenRouter ids: tencent/hy4, z-ai/glm-5.3-flashx-legal, deepseek/deepseek-v4.1-flash - Date: November 5, 2026. Temperature zero, one call per task, no retry to fish for a better answer - Archive: 44 invoices, 48 emails, 400 support tickets, 6 monthly service reports, 12 sets of minutes and 3 amendments. Measured at 61,650 input tokens by the API - Tasks: unsettled invoices and outstanding total; rate-card audit; renewal notice deadline; service credit for the month that missed the availability target - Scoring: seven fields with a fixed answer key computed from the archive before the first call, not by taste - First round used a 900-token output budget. Where a model returned an empty message we re-ran the same task at 4,500 and then 12,000 tokens, and the video says which round counted for each model. DeepSeek needed 9,565 thinking tokens on the first task; Hy4 returned nothing at any of the three budgets and only answered with its reasoning mode disabled - Cost per pass in the scoreboard is list price for the four tasks at 61,650 input tokens plus the output tokens each model actually wrote. DeepSeek's pass is cheap because it writes short answers, not because it thinks less: its nine thousand five hundred thinking tokens are billed on top - Total billed for this episode: US$0.53 across 30 calls, including the discarded rounds. Every call is in our cost ledger - We did not test the one million token claim. We tested a folder size a finance team actually has Chapters: 0:00 61,650 tokens of invoices 0:14 How we tested: three 1M-token models 1:24 Task 1: which invoices are still open 1:53 Task 1 answer: 12 open, $165,576.36 2:32 Task 2: audit the rate card 3:01 Task 2 answer: overcharge and a phantom fee 3:19 Task 3: 60 days or 90 days 3:47 Task 4: the service credit 4:22 Hy4 returned nothing, nine times 5:01 Scoreboard: fields correct and the bill 5:28 Verdict: one hire, one hype, one catch 6:12 Next episode Sources: - Tencent, "Tencent Launches Hy4, Joining the Ranks of the World's Leading Open-Source AI Models", November 4, 2026: https://www.tencent.com/tencent-launches-hy4-joining-the-ranks-of-the-worlds-leading-open-source-ai-models/ (accessed November 5, 2026) - SiliconANGLE, "Tencent releases Hy4, a powerful frontier LLM introducing visual understanding and computer use", November 4, 2026: https://siliconangle.com/2026/11/04/tencent-releases-hy4-powerful-frontier-llm-introducing-visual-understanding-and-computer-use/ (accessed November 5, 2026) - Z.ai documentation, GLM-5.3-FlashX and GLM-5.3-FlashX-Legal pricing and model description: https://docs.z.ai/guides/vlm/glm-5.3-flash (accessed November 5, 2026) - OpenRouter model pages and prices, accessed November 5, 2026: https://openrouter.ai/tencent/hy4 · https://openrouter.ai/z-ai/glm-5.3-flashx-legal · https://openrouter.ai/deepseek/deepseek-v4.1-flash Narration uses an AI voice. Subscribe for weekly AI tool verdicts. #AItools #HypeOrHire #LLM #Accounting #ContractReview

Selected Topics

  • Dogecoin whale activity
    Dogecoin whale activity
    Get the latest insights into Dogecoin whale activities with our comprehensive analysis. Discover trends, patterns, and the impact of these whales on the Dogecoin market. Stay informed with our expert analysis and stay ahead in your cryptocurrency journey.
  • Dogecoin Mining
    Dogecoin Mining
    Dogecoin mining is the process of adding new blocks of transactions to the Dogecoin blockchain. Miners are rewarded with new Dogecoin for their work. This topic provides articles related to Dogecoin mining, including how to mine Dogecoin, the best mining hardware and software, and the profitability of Dogecoin mining.
  • Spacex Starship Launch
    Spacex Starship Launch
    This topic provides articles related to SpaceX Starship launches, including launch dates, mission details, and launch status. Stay up to date on the latest SpaceX Starship launches with this informative and comprehensive resource.
  • King of Memes: Dogecoin
    King of Memes: Dogecoin
    This topic provides articles related to the most popular memes, including "The King of Memes: Dogecoin." Memecoin has become a dominant player in the crypto space. These digital assets are popular for a variety of reasons. They drive the most innovative aspects of blockchain.