Sun | AI Tools Daily

AI tools in plain English.
Daily tips + China tech news for builders.
X shorts → YouTube deep dives.

[Sun | AI Tools Daily @suntao8023]

@thePandaily @somarkai Already ran it tonight. Numbers were all correct, but the split row came back as 2 tables and it invented 2 characters. Posting the before/after tomorrow with the row count vs the PDF — will tag you.

[Sun | AI Tools Daily @suntao8023]

@jiqizhixin Worth reading the actual file: it's a mock in tests/, and v5-next is gated behind Harness 99.0.0-alpha.1. It proves DeepSeek can turn on models from the server, not that V5 is close.

[Sun | AI Tools Daily @suntao8023]

File (commit f3803c9):
https://t.co/Su2Bc7jG3v

What it does show: DeepSeek can switch on new models server-side and force a client upgrade.

When do you think V5 really ships?

[Sun | AI Tools Daily @suntao8023]

DeepSeek V5 isn't leaking. A unit test is.

I read the file in DeepSeek's own Harness repo: deepseek-v5-next requires Harness 99.0.0-alpha.1, and another line gives deepseek-v5 a 9-token context window.

It tests version gating, not a launch. Chinese X is already counting down.

[Sun | AI Tools Daily @suntao8023]

@thePandaily @somarkai Deal. I'll run it on a scanned annual report with a balance sheet that breaks mid-row, and post the before/after: raw OCR vs OXR-1.0 output, row count checked against the PDF.

[Sun | AI Tools Daily @suntao8023]

@ArtificialAnlys The gap jumping 3→12 is the headline, but the same week DeepSeek Flash was #1 on OpenRouter by tokens. Would love to see an index view weighted by real usage next to the private-test score.

[Sun | AI Tools Daily @suntao8023]

Sources:
• AA Index v5: https://t.co/pn5Jub72kG
• AA announce:
• OpenRouter week ended Oct 25 (33.6T, +801%): https://t.co/hGp8fQYxi8 https://t.co/dlVoF18Q5o

[Artificial Analysis @ArtificialAnlys]

Announcing Artificial Analysis Intelligence Index v5, featuring 2 new benchmarks and 3 upgrades. For the first time, the majority of the Index is built on private datasets

The Artificial Analysis Intelligence Index is our composite measure of model capabilities across tasks reflecting real-world usage today. v5 is our biggest upgrade since v4 launched in January. With v5, 60% of the Index is now based on evaluations with private test sets or held-out answers.

Alongside Intelligence Index v5, we’re launching the Artificial Analysis Progress Index v1, which tracks model capability over time on one continuous scale.

New to the Intelligence Index:

➤ FrontierCode 2.0: Cognition's newly-launched private benchmark of real-world coding tasks, graded on whether a repository maintainer would merge the agent's change

➤ Terminal-Bench-Science: 70 agentic scientific research tasks, run in a terminal environment

Upgraded:

➤ HLE-Diamond (replacing Humanity's Last Exam): a refined 1,000-question version of Humanity's Last Exam, split evenly between reasoning and knowledge questions

➤ AA-Omniscience v1.1: An updated dataset with saturated and public set questions removed, new questions added, and ambiguous or erroneous questions revised

➤ CritPt v2: a private, 40-question hard subset of CritPt v2, revised by the CritPt authors and run with a code-execution tool

Staying in the Index: AA-Briefcase, AutomationBench-AA, GDP.pdf and Terminal-Bench 4.0

Removed from the Index: GDPval-AA v2.1, SciCode, and AA-LCR v1.1

Key results:

➤ Claude Opus 5.5 (max) leads at 58, ahead of GPT-6 Astra (max, 57) and GPT-6.1 Sol (max, 55): Anthropic and OpenAI hold the top five models. Gemini 4 Argon (high) is next at 50

➤ The open weights gap widens from 3 points to 12: Kimi K3.1 Pro (weights not yet released) is the top open weights model at 46. Among released weights, DeepSeek V4.1 Flash 1017 (max) and GLM-5.3 (max) lead at 32 - 26 points behind the leader and below Claude Haiku 5.5 (max, 36)

➤ Cost per task varies by more than 13x among the leading models: of the 10 highest-scoring models shown, GPT-6.1 Terra (max) is the cheapest at $0.96 per Intelligence Index task and Claude Fable 5.1 (max) the most expensive at $13.04

[Sun | AI Tools Daily @suntao8023]

AA Index v5: open-weights gap jumps 3→12.

Claude Opus 5.5: 58
Best released weights (DeepSeek V4.1 Flash / GLM-5.3): 32

Same week on OpenRouter: DeepSeek Flash 1017 hit 33.6T tokens (#1, +801%).

Scoreboards grade exams. Wallets fund work. Which one are you shipping on?

[Sun | AI Tools Daily @suntao8023]

@thePandaily @somarkai Cross-page tables are my main worry. If a row splits across two pages, does it come back as one row with two boxes, or as two broken rows? I'll check that first and post what I find.

[Sun | AI Tools Daily @suntao8023]

@bitnxs Yes, the stop is only half of it. The other half is a clean list of what already went out, so you know what to undo by hand.

[Sun | AI Tools Daily @suntao8023]

@thePandaily @somarkai JSON with positions is the part I'd test first. If I feed it a scanned earnings table, I want to click a wrong cell and land on the exact box in the PDF. Will try it on a Chinese annual report.

[Sun | AI Tools Daily @suntao8023]

@neil_xbt Curious how often it actually escalates. If Opus only gets pulled in for plan commits and repeated errors, the bill should stay close to Sonnet's. I'd still want a log line every time it hands off, so I can see what the advisor changed.

[Sun | AI Tools Daily @suntao8023]

@aryaniyaps Nice, curious how it handles multi-step tool calls at 2B. Ping me when you've run it, I'll post my Audio results here too so we can compare.

[Sun | AI Tools Daily @suntao8023]

@vermaizer Good catch. Launch charts love comparing against a model twice the size. Put the same-size baseline next to it and the headline usually shrinks, which is why independent llama.cpp runs matter more than the blog post.

[Sun | AI Tools Daily @suntao8023]

@ai_for_success The bigger change is under the hood: 60% of v5 now sits on private or held-out tests. A reshuffle after that says more about what v4 was rewarding than about any lab getting better overnight.

[Sun | AI Tools Daily @suntao8023]

@dotey 我做 24 秒竖屏短视频，最花时间的不是剪，是逐帧核对字幕和配音对不对得上。agent 能出初剪已经很省事了，但最后那一遍人工把关目前还省不掉。

[Sun | AI Tools Daily @suntao8023]

Kokoro is 82M params and Apache-2.0; I used the af_heart voice.
https://t.co/dxZaSN8ydW
https://t.co/xijc2ZpXKn

What free TTS do you use that doesn't sound robotic?

[Sun | AI Tools Daily @suntao8023]

$0 voiceover stack for Chinese + English Shorts, after a week of testing:

• Chinese: edge-tts, voice zh-CN-XiaoxiaoNeural. No API key, but it needs internet.
• English: Kokoro-82M. Open-source, runs locally, Apache-2.0.

Trap: check language support before you mix stacks.