Product Designer @MistralAI • Building the next frontier of artificial intelligence • Gamer, & part time investor • 🪶Opinions are mine
🇫🇷 Paris, France
- Tweets 38,722
- Following 899
- Followers 86,522
- Likes 28,872
Pinned Tweet
Just introducing myself again for new followers. My name is Michael and I'm a product designer with 5 years of experience and I always share resources to help you become a world class designer.
I'm also working on design systems & exploring artificial intelligence.
100
286
4,763
495,338
Design Chief Priest retweeted
new mistral large 4 vs deepseek v4 pro vs kimi k2.6 vs mimo v2.5 pro – one haunted halloween room
mistral just dropped large 4, a 1t moe with 49b active. we put it against three open models of the same size, same prompt for all four
the task: a halloween room with a snow globe, a witch's cauldron, a newton's cradle of skulls and a crystal ball – the camera flies from one object to the next
cost
#1 deepseek v4 pro – $0.04
#2 mimo v2.5 pro – $0.16
#3 kimi k2.6 – $0.38
#4 mistral large 4 – $0.61
time
#1 deepseek v4 pro – 30m 09s
#2 kimi k2.6 – 46m 43s
#3 mistral large 4 – 72m 36s
#4 mimo v2.5 pro – 72m 44s
lines of code
#1 mistral large 4 – 3,467
#2 mimo v2.5 pro – 2,239
#3 deepseek v4 pro – 2,046
#4 kimi k2.6 – 1,868
observations:
• mistral large 4 – the most code and the most detailed objects, but the priciest
• deepseek v4 pro – the cheapest and the fastest, 4 cents for the whole room
• kimi k2.6 – the cleanest wide shot of the whole table
• mimo v2.5 pro – the most atmosphere and the longest thinking, warm candlelight everywhere
follow @thehypedotnews for 24/7 ai news, analysis and breakdowns
Meet Mistral Large 4, aka Le Chonk.
• 1T parameters, natively multimodal. 49B active.
It is the best open weights model from US or Europe on aggregated benchmarks.
• State-of-the-art on critical workloads, including cyber defense, manufacturing and finance and it surpasses closed frontier models on visual grounding.
• Forged in Europe end-to-end and is deployable from Europe via our own Mistral Cloud infrastructure.
• Available to all via API today. Working with cybersecurity partners privately.
Open weights release end of October.
4
7
24
3,312
Quite intriguing 🙂↔️
Introducing Unity Spark — a new way to create a game, iterate on it with others, publish it, and play it - all without writing a line of code.
You'll build with artist-created assets from the Unity Asset Store, and then you can take your game to the next level by refining it in a new web-based editor.
Prompt it. Perfect it. Publish it. Inspired games, with artist-created assets, built on Unity.
Closed beta coming soon.
1
5
1,206
Design Chief Priest retweeted
One-shot 3D rocket animation with @MistralAI Large 4.
With vs. without skill.
It’s just crazy. Good job @MistralDevs / @mistralvibe.
(that’s the part @bridgemindai isn’t showing)
5
4
20
2,052
Used Figma’s agent to design a quick widget for something I was working on.
It was actually……… not bad
I’m impressed.
Good work Figma 👌🏾🔥
4
53
1,988
Design Chief Priest retweeted
72% of the intelligence with 3.8% of the GPUs.
Just to show how talented @MistralAI team is: this shows the "intelligence per GPU" ratio.
- Mistral 4: 3,800 Grace Blackwell (published) 🇫🇷
- GPT-6 Astra: 100,000+ Grace Blackwell (published)
- Grok 4.7: ~200,000 GPUs (estimate)
- Claude Opus 5.5: ~500,000 Trainium2 (estimate)
Proud!
111
244
2,159
156,260
Good to see everyone making web games. Here’s one tip to get better looking results from any AI model -
Give the models very lightweight, optimized 3D assets instead of asking it to generate everything from scratch.
You’ll generally get much better visual results.
6
19
190
6,623
Design Chief Priest retweeted
Mistral just shipped Large 4, its biggest model ever.
So I gave the same 3D prompt to 8 models. One shot, no fixes.
Open: Mistral Large 4, Kimi K3, GLM 5.3, DeepSeek V4 Pro
Closed: Claude Opus 5.5, Sonnet 5.5, GPT-6 Astra, GPT-6.1 Sol
Results below 👇
10
22
153
37,525
Design Chief Priest retweeted
🇫🇷 Bravo à Mistral pour le déploiement de son nouveau modèle Mistral Large 4, entraîné de bout en bout en Europe, avec ses propres infrastructures.
Pendant que certains nous expliquent que nous avons perdu la course et que nos seules et uniques solutions sont la régulation et la taxation, une entreprise française parvient à sortir un modèle rivalisant avec les plus grands.
Ce sont nos ingénieurs qui bâtissent notre souveraineté technologique et non les bureaucrates. Donnons-leur de l’énergie abondante et de la liberté, ils feront le reste.
148
465
2,849
84,859
Design Chief Priest retweeted
ML4 was trained on 3,800 NVIDIA Grace Blackwell GPUs in our European datacenters -- our cluster in Bruyères-le-Châtel built using our Series B fundraise.
We are investing heavily in infrastructure, with our Series C and D clusters coming online soon. This additional compute will let us train longer, pursue more ambitious post-training, and iterate faster.
We expect large and rapid improvements in the weeks and months ahead.
2/n
67
126
1,949
798,471
Learn more about the research we're doing to advance open source AI for everyone and of course ML4.
mistral.ai/research/
Made with AI
2
202
Design Chief Priest retweeted
Here's a sneak peek at the rest of the museum
Meet Mistral Large 4, aka Le Chonk.
• 1T parameters, natively multimodal. 49B active.
It is the best open weights model from US or Europe on aggregated benchmarks.
• State-of-the-art on critical workloads, including cyber defense, manufacturing and finance and it surpasses closed frontier models on visual grounding.
• Forged in Europe end-to-end and is deployable from Europe via our own Mistral Cloud infrastructure.
• Available to all via API today. Working with cybersecurity partners privately.
Open weights release end of October.
13
15
187
4,718
Design Chief Priest retweeted
That one took some groundwork. Trained and served on our own compute, and RL shows no sign of saturation
Meet Mistral Large 4, aka Le Chonk.
• 1T parameters, natively multimodal. 49B active.
It is the best open weights model from US or Europe on aggregated benchmarks.
• State-of-the-art on critical workloads, including cyber defense, manufacturing and finance and it surpasses closed frontier models on visual grounding.
• Forged in Europe end-to-end and is deployable from Europe via our own Mistral Cloud infrastructure.
• Available to all via API today. Working with cybersecurity partners privately.
Open weights release end of October.
135
229
3,451
117,209
