Mixtral
product on 9 shows · 6 statements across 4 episodes · said 28 times in 21 episodes since 2023
Latent Space 14 
the a16z Podcast 5 
the MAD Podcast 3 
No Priors 2 
BG2 Pod 1 
the Official SaaStr Podcast 1 
All-In 1 
TBPN 1 
20VC
Mentions by year, every show
tap a year for its mentions
Latent Space 14the a16z Podcast 5the MAD Podcast 3No Priors 2All-In 1BG2 Pod 1the Official SaaStr Podcast 1TBPN 1
2026 3 mentions in 3 episodes 1 per episode
2025 5 mentions in 4 episodes 1 per episode
- Building Jamba 3B: the tiny Hybrid Transformer State Space Reasoning Model - Barak Lenz, CTO of AI21
- Why RL Won — Kyle Corbitt, OpenPipe (acq. CoreWeave)
- DeepSeek, Reasoning Models, and the Future of LLMs
- DeepSeek V3, SGLang, and the state of Open Model Inference in 2025 (Quantization, MoEs, Pricing)
- every mention in 2025, scene by scene →
2024 16 mentions in 13 episodes 1 per episode
- A Brief History of the Open Source AI Hacker - with Ben Firshman of Replicate
- The Four Wars of the AI Stack - Dec 2023 Recap
- AI Semiconductor Landscape feat. Dylan Patel | BG2 w/ Bill Gurley & Brad Gerstner · Bg2 Pod
- The State of AI Startups in 2024 [LS Live @ NeurIPS]
- Making AI Work: Fine-Tuning, Inference, Memory | Sharon Zhou, CEO, Lamini
- The 10,000x Yolo Researcher Metagame — with Yi Tay of Reka
- No Priors Ep. 63 | With Sarah Guo and Elad Gil
- How Shopify Implements AI Across Sales and Product with the Head of AI at Shopify
- 5 more episodes that year, every mention in 2024 →
2023 4 mentions in 1 episode
every mention on every show, scene by scene, with the transcript →
6 statements about Mixtral, every show
Ethan He: NVIDIA Cosmos uses a 7B video model with a larger LLM rewriter
“I think in in Cosmos, we use Lama or we use mix, mix through. And the Cosmos video model itself is only seven B, and the model, the language model is a prompt rewriter. It's bigger than that.”
He: Mixtral's top-k before softmax routing hurts MoE upcycling performance
“We actually found the mix-throughs approach didn't work as well as expected, because the original model, the original switch transformer from Google uses a softmax and topk for a reason. And because of upcycling, if you switch to topk, then softmax, it actuall…”
Enterprises prototype with GPT-4 but shift to cheaper self-hosted models for production
“We see quite a bit of usage patterns where people would start GPT-IV for design and then decide to move to Essentially something cheaper, like nixtral self-hosted or nixtral self-hosted when moving to production.”
Mensch: Mixtral has 46B total parameters but executes 12B per token
“You have eight experts per layer and you execute only two of them. So what it means at the end of the day is that you have a lot of parameters on your model. You have forty six billion parameters, but the thing is that the number of parameters that you execute…”
Mensch: Mixtral matches Llama 2 70B performance at one-sixth the cost
“Mixtral is actually on par with Lama-to-seven TB while being approximately six times cheaper or six times faster for the same price.”