Training, testing & serving AI models.
- Tweets 851
- Following 349
- Followers 952
- Likes 514
Pinned Tweet
Gemini 4 Pro (in arena) vs Claude Opus 5.5 on a 3D floatplane physics simulation
>Gemini 4 pro is completely outperforms Claude opus 5.5 here
> When it officially drops, it’s going to raise the bar completely.
77
35
959
194,729
Claude Opus 5.5 vs. DeepSeek V4.1 Flash vs. MiniMax M3.1 Flash (wait till the end)
>DeepSeek mogged MiniMax 3.1 Flash
>People are saying DeepSeek V4.1 Flash is close to Opus 5.5 in 3D modeling, but I don't agree.
2
20
2,102
I gave Claude Opus 5.5 a prompt to build a villa from scratch. Here’s what it made.
>Is Opus 5.5 really nerfed?
2
7
538
Gemini 4 Carbon is already here in early access .
>Partners are testing it on LMArena and the full release is very close.
>It’s clearly better than Argon at 3D… just still has a bad UI and is slower (but way less lazy).
Gemini 4 Carbon EAP has officially started. Early access has been given to partners and it's being tested in LMArena. The release is very close.
It's noticeably better than Gemini 4 Argon at 3D, though frontend/UI is still bad. It's also slower but less lazy.
4
310
Claude Fable 5.5 is very close to release.
>Fable 5.5 is set to compete with Opus 5.5, rather than with others.
>As one employee said, Gemini 4 Carbon is very close to Opus 5.5.
>Here, Argon is out of the competition against Fable 5.5.
1
2
259
MiniMax M3.1 Flash benchmarks beats DeepSeek and Kimi
>let’s see whether these benchmarks translate into real performance or just benchmaxxed
1
9
599
The worst-kept secret in AI is finally official: Space Bunny = MiniMax-M3.1-Flash
>MiniMax-M3.1-Flash is officially leaving preview after dominating OpenRouter & OpenCode with 1M context.This one’s going to hit different.
9
755
Claude Opus 5.5 vs. DeepSeek V4.1 Flash-1017 vs. Kimi K3
>DeepSeek completely mogged Kimi K3
>let's be honest,nothing touches Opus 5.5 right now.
27
5
248
28,753
According to some people:
Today: DeepSeek V4.1 beats Opus 5.5
Tomorrow: DeepSeek V4.1 is AGI
Day after: DeepSeek V4.1 is ASI
3
6
844
DeepSeek V4.1 Flash-1017 vs Opus 5.5 vs V4.1 Flash on 3D horse simulation.
>V4.1 Flash-1017 is way better than V4.1 Flash.
>Seen many videos where Flash-1017 matches Opus 5.5.
Here’s what I got:
23
10
309
55,193
These are the outputs I got from Fable 5.5 when it was routed from Opus 5.5.
>High chance Fable 5.5 release this week.
7
2
34
2,919
🚨 Three new OpenAI models just leaked:
aurora-red-alpha
aurora-green-alpha
aurora-blue-alpha.
> aurora-red-alpha is already throwing a 503 capacity error in model testing
>while green and blue are returning 404s.
1
1
11
906
My current top 5 AI models ranking
1.Claude Opus 5.5
2.Claude Opus 5.5
3.Claude Opus 5.5
4.Claude Opus 5.5
5.Claude Opus 5.5
Btw, i am not big fan of Claude
1
4
335
DeepSeek V4.1 Flash-1017 is outperforming GPT-5.6 Sol and Claude Opus 5 on coding and agent benchmarks while costing a fraction of the price.
>Flash: $0.15/$0.60 (off-peak) vs Sol $4/$20 and Opus $5/$25.
>Price/performance is wild.
Major update: DeepSeek V4.1-Flash-1017 is here.
Clear improvements over the previous version on key agent and coding benchmarks.
1
172
Claude’s 1.5x faster in chat now. That crucial wait before the first token just got cut from 3.36s to 2.17s. Time saved from now on.
4
421
Major update: DeepSeek V4.1-Flash-1017 is here.
Clear improvements over the previous version on key agent and coding benchmarks.
🚀 DeepSeek-V4.1-Flash-1017 is here!
Stronger benchmark performance, faster task completion, and shorter outputs than the previous version—a Pareto improvement in performance and cost. Now live on our website, app, and API.
🛠️ Optimized for @DeepSeekHarness, Tencent CodeBuddy, WorkBuddy, and more. Upgrade to DeepSeek Harness v0.2.1-rc.1 or later for the best experience with the new model.
1/4
1
427
