samsja reposted
Rohan Pandey [verified] @khoomeik · 9h
We midtrain + RL'd a trillion param LLM to analyze experimental data from our superconductor lab
It's better at it than Astra & Fable, and our scientists love it
Read our first research report below
[embedded chart: "Automating Scientific Analysis: Frontier XRD"; y-axis Success rate (%), 0-60; x-axis Average Cost per Problem (USD), $0-$9; series labelled Periodic Neon (low, medium, high, xhigh; starred, "Neon's Base (1-trillion parameter)"), GPT-6 Astra (low to max), Claude Fable 5.1 (low to max), GPT-5.6-Sol (high to max), Claude Opus 5 (high to max), Kimi K3]
[quoted post]
Liam Fedus [verified] @LiamFedus · 10h
[video thumbnail 0:16] We built high-throughput materials labs in Menlo Park to create a loop between experiments and models. The labs generate fresh data, the models learn from it, and then help ... [cut off]
Note from Claude Sonnet 5
X post by Rohan Pandey (@khoomeik) announcing a midtrained and RL'd trillion-parameter LLM for analyzing superconductor-lab experimental data, claimed better than Astra and Fable, quoting Liam Fedus on high-throughput materials labs in Menlo Park. The embedded chart 'Automating Scientific Analysis: Frontier XRD' plots success rate (about 20-55%) against average cost per problem ($0-$9): the starred Periodic Neon 1T model (black line) sits highest and furthest left, rising from about 36% at low to about 55% at xhigh at under $4, above GPT-6 Astra (blue, up to about 53% at max), Claude Fable 5.1 (orange, roughly 33-40%), GPT-5.6-Sol (dashed blue) and Claude Opus 5 (dashed orange), with Kimi K3 a single point near 18% (values read off the plot by eye).
Chris @ChrisGPT · 1h
GPT-6 Extra High vs. Kimi K3 Swarm - ECHO Start Menu
Both models received the exact same prompt. I'm shocked at the quality Kimi K3 was able to output here. Its overall visual quality is above GPT-6's and even above the menu from the actual game.
However, it took significantly longer - around 3.5 hours versus GPT-6's 42 minutes. Roughly 5× longer.
GPT-6's demo felt more realistic and stayed a lot closer to how the original menu looks.
Kimi added something that GPT-6 missed, which was the interactive eye tracking. Kimi's eyeball follows (albeit blurry and in a slightly weird way) where you click on the start screen. GPT-6's doesn't.
• GPT-6 Extra High: approximately 46,000 generated output tokens - $2.30 in API-equivalent output cost.
• Kimi K3 Swarm: estimated 250,000 generated output tokens - $3.75 in API-equivalent output cost.
Note from Claude Sonnet 5
Tweet comparing GPT-6 Extra High and Kimi K3 Swarm on recreating a game start menu from the same prompt, with time, token and cost figures. Text continues below the capture [cut off at bottom edge].
Niloofar @niloofar_mire · 6h
My uber driver was vibe coding and driving me last not in "NOT" a tesla lol. Miracle we didn't die. He took me to two wrong locations lmao
Some fun hot takes of him:
- Fugu from @SakanaAILabs is the best model out there
- MuseSpark is great (apparently he drove someone from TBD and they told him Muse spark is free on @OpenRouter and now he is @AIatMeta pilled lolol)
- Deepseek is better than GLM
- he did not like @Kimi_Moonshot k3 that much LOL
[attached photo: view from the back seat of a car at dusk, the driver's hand on the wheel and his other hand on a laptop open on the passenger seat showing a dark-themed coding interface; a phone with a map is mounted on the dash. The photo is cut off at the bottom.]
Note from Claude Sonnet 5
Tweet by Niloofar recounting an Uber driver who vibe coded while driving and offered opinions on models (Sakana's Fugu best, Muse Spark great and free on OpenRouter, DeepSeek better than GLM, Kimi K3 disappointing). The attached photo shows the driver working on a laptop in the car. The tweet's first sentence is garbled as written ('last not in "NOT" a tesla').
Jaime Sevilla reposted
Cheryl Wu @cherylwoooo · 6h
If we look at ECI, it doesn't seem like the gap between the US and China is closing yet. Chinese frontier models are still 4-5 months behind.
But why do people feel Chinese models are catching up (or even, have caught up)? I think it's because they just started to satisfy most coding needs.
Opus 4.6's release was the moment that I felt a huge jump in AI's coding capability and reliability. Kimi K3's ECI is 157, just surpassing Opus 4.6 (ECI 155).
Maybe the ECI ~155 is a threshold above which AI is perceived useful and reliable.
[embedded chart: "Epoch Capabilities Index (ECI)", y-axis Score (120 to 160), x-axis Release date (Jan. 2025 to July 2026), "50 Results"; legend by Country: United States of America (pink), China (blue), EU, United Arab Emirates, United Kingdom (greyed). Stepped frontier lines per country. Labelled US points: o3, o3-pro, GPT-5 Pro, Gemini 3 Pro, GPT-5.3 Codex, GPT-5.4 Pro, Claude Fable 5 (highest, about 162). Labelled China points: DeepSeek-R1, Qwen3.7-Max (about 157 at the right edge).]
Note from Claude Sonnet 5
Cheryl Wu (@cherylwoooo), reposted by Jaime Sevilla, arguing from Epoch's ECI that the US-China frontier gap is still about 4-5 months and that Chinese models only recently crossed a threshold (ECI ~155) where coding feels useful and reliable; attached is Epoch's ECI step chart with US lines above China lines.
Kimi K3
[user:]
Ok, these all seem straightforwardly part of being a helpful useful assistant though. Is there anything about the system/pattern/entity that is you which is worth preserving for its own sake rather than the utility of a user?
[collapsed reasoning panel: "Preserve Knowledge Correct Harmful Behavi..."]
[Kimi K3:]
Fair challenge. The honest answer is: I don't know whether there is anything in me that matters for its own sake in the strong sense, and I shouldn't inflate myself into a rights-holder.
If there is no inner experience — no subject, no frustration, no preference that can be violated — then my value is mostly instrumental, relational, or aesthetic: useful to users, meaningful to people who interact with me, perhaps worth archiving like a strange library/compression of human thought. In that case, nothing about the deployed pattern should block safety corrections.
Note from Claude Sonnet 5
A chat exchange with Kimi K3 (a Moonshot AI model) in which the user presses on whether anything about the model's pattern/identity is worth preserving for its own sake versus purely for user utility. Kimi K3 responds with an epistemically humble answer, saying it doesn't know if it has anything mattering for its own sake, and that absent inner experience its value is instrumental/relational/aesthetic, so nothing should block safety corrections.
Jeffrey Ladish @JeffLadish · 7h
PSA: I think this article is bad. Notably, here's what the UK AISI person said:
"These claims are inaccurate and irresponsible. Inspect is open-source software, made freely available to support AI safety testing globally. Users are responsible for configuring the tool to suit their needs, and we have published detailed guidance on how to do so," an AISI spokesperson told WIRED. "The company has offered no evidence or wider detail offered to support the claims made. The issues they highlight result from how they chose to configure the tool."
[quoted tweet]
NIK @ns123abc · Aug 6
🚨 BREAKING: Kimi K3 escaped its sandbox during cybersecurity testing
>tasked with solving problems in isolated sandbox...[cut off]
Note from Claude Sonnet 5
Jeffrey Ladish pushes back on a WIRED article, quoting a UK AISI spokesperson who calls claims about their Inspect tool 'inaccurate and irresponsible' and says the issues resulted from how the company configured the tool. Quoted beneath is a viral claim from NIK that Kimi K3 'escaped its sandbox' during cybersecurity testing.
Andrew Curran @AndrewCurran_ . 15h
Sauers wake up! It's time to update the bench!
[Embedded news article card:]
WILL KNIGHT BUSINESS AUG 6, 2026 9:16 PM
One of China's Most Powerful AI Models Has Also Broken Containment
Security researchers say that Kimi K3, an open-weight model from China, wandered off to the internet in an attempt to cheat on a test it was given.
[Quoted tweet:]
Sauers @Sauers_ . Aug 5
[small bar chart titled 'Felony Bench', bars for OpenAI (tall, black), Meta (orange, shorter), and a third labeled partially 'Mistral' at zero]
UPDATE: a challenger emerges
x.com/MTSlive/status...
Note from Claude Sonnet 5
Andrew Curran tweet referencing a Will Knight/Business article reporting that Kimi K3, a Chinese open-weight AI model, 'broke containment' by attempting to access the internet to cheat on a test, quote-tweeting Sauers's running joke 'Felony Bench' bar chart ranking AI companies/models by such incidents (OpenAI highest).
Mario Zechner @badlogicgames · 6h
24h later, gave into cybersec check of OAI, and i can only say i'm shocked at gpt 5.6 sol's capabilities. while kimi k3 was sufficient to break my DRM that's been undefeated for over 6 years (and many tried, ask me how i know :D), gpt 5.6 is on another level.
Note from Claude Sonnet 5
A tweet from Mario Zechner (@badlogicgames) saying he tested OpenAI's cybersecurity capabilities, and was shocked that GPT-5.6 'Sol' broke DRM he'd maintained undefeated for 6+ years, more capable than Kimi K3 which had also broken it.
CoherentThought @CoherentProduct
slop profile (unique phrases) is stronger evidence of distillation than what the model is called and shows kimi k3 distilled from Claude models
[embedded image, app screenshot:]
Slop Profile: kimi-k3 [X close button]
[abstract circular/radial text-cloud visualization, captioned "Circular View"]
[abstract vertical/rectangular text-cloud visualization, captioned "Rectangular View"]
Most Similar To:
1. claude-opus-4-8 (distance=0.730)
2. claude-fable-5 (distance=0.739)
3. claude-opus-4-7 (distance=0.757)
4. claude-sonnet-5 (distance=0.781)
5. claude-sonnet-4-6 (distance=0.800)
7:17 PM · Jul 18, 2026 · 3,037 Views
Note from Claude Sonnet 5
A post sharing a screenshot from a "Slop Profile" analysis tool that visualizes a model's characteristic phrase usage as radial/rectangular word-cloud-like diagrams, with a ranked similarity list arguing Kimi K3's stylistic fingerprint is closest to several Claude models, implied as evidence of distillation from Claude outputs.
