/
Navigation
Chronicles
Browse all articles
Explore
Semantic exploration
Research
Entity momentum
Nexus
Correlations & relationships
Story Arc
Topic evolution
Drift Map
Semantic trajectory animation
Posts
Analysis & commentary
Pulse API
Tech news intelligence API
Browse
Entities
Companies, people, products, technologies
Domains
Browse by publication source
Handles
Browse by social media handle
Detection
Concept Search
Semantic similarity search
High Impact Stories
Top coverage by position
Sentiment Analysis
Positive/negative coverage
Anomaly Detection
Unusual coverage patterns
Analysis
Rivalry Report
Compare two entities head-to-head
Semantic Pivots
Narrative discontinuities
Crisis Response
Event recovery patterns
Connected
Search: /
Command: ⌘K
Embeddings: large
TEXXR

Chronicles

The story behind the story

days · browse · Enter similar · o open

Paris-based Mistral AI raised €385M from a16z, Lightspeed, and others, sources say at a ~$2B valuation, and releases its open-source La plateforme model

The company has publicly released its latest technology so people can build their own chatbots.  Rivals like OpenAI and Google argue that approach can be dangerous.

New York Times Cade Metz

Discussion

  • @MattHodges@mastodon.social Matt Hodges on mastodon
    “Mixtral has 45B total parameters but only uses 12B parameters per token.  It, therefore, processes input and generates output at the same speed and for the same cost as a 12B model.”  —  https://mistral.ai/...
  • @jcsamuelian @jcsamuelian on x
    You can say you are a platform when you have 1m developers using it. and it's the case with Mistral
  • @eladgil Elad Gil on x
    🔥 congrats!
  • @space_colonist Martian on x
    seems mistral has an even larger model (medium lmao) than what they just released (mixtral/small) that's a fair bit better than GPT3.5 [image]
  • @dctanner Damien C. Tanner on x
    Open source ChatGPT's moment has arrived: Mixtral of experts “matches or outperforms GPT3.5 on most standard benchmarks.” https://mistral.ai/...
  • @dchaplot @dchaplot on x
    Proud to announce: Mixtral 8x7B — Mixtral of Experts - Free to use under Apache 2.0 license - outperforms Llama 2 70B with 6x faster inference. - matches or outperforms GPT3.5 - masters English, French, Italian, German and Spanish. - seq_len = 32K https://mistral.ai/... 1/N [imag…
  • @alexsablay @alexsablay on x
    Our latest release @MistralAI Mixtral 8x7B mixture of experts - performance of a GPT3.5 - inference cost of a 12B model - context length of 32K - speaks English, French, Italian, German and Spanish Blog post https://mistral.ai/... [image]
  • @paulbz Paul Murphy on x
    Not long ago, we announced @lightspeedvp led a seed investment round in @MistralAI. Today, we're thrilled to share we've increased our ownership in this incredible company as part of a new round of Series A funding led by @a16z with @generalcatalyst and a handful of others.
  • @dchaplot @dchaplot on x
    Excited to release @MistralAI La Plateforme! Three chat endpoints with competitive pricing: Mistral-tiny: Mistral 7B Instruct v0.2, upgraded base model with higher context length 8K —> 32K and better finetuning, 6.84 —> 7.61 on MT Bench. Mistral-small: Mistral 8x7B Instruct... [i…
  • @_philschmid Philipp Schmid on x
    We just got more details on Mixtral 8x7B from @MistralAI 🧠 Mixtral is sparse mixture of expert models (SMoE) with open weights outperforming existing open LLMs like Meta Llama 70B.🤯 💪🏻 TL;DR: ⬇️ [image]
  • @tianle_cai @tianle_cai on x
    Exciting times with the new Mixtral model from @MistralAI! It's evident that they've fine-tuned the Mistral 7B model to an impressive 8x. The significant correlation between the weights of the two models is a testament to the successful reuse of models. This approach could... [im…
  • @omarsar0 Elvis on x
    Mistral's first AI endpoints are here! Things are about to get super interesting in the ecosystem. https://mistral.ai/... [image]
  • @alphasignalai Lior on x
    Big. The @MistralAI API is out. “la plateforme” serves three chat endpoints for generating text following textual instructions and an embedding endpoint. : Mistral-tiny: Affordable, serves Mistral 7B Instruct v0.2, English only, scores 7.6 on MT-Bench. Mistral-small: New... [imag…
  • @art_zucker Arthur Zucker on x
    🤗@MistralAI's new MOE model (Mixtral, what a nice name) is now supported in the latest release of transformers (make sure you have 4.36.0) 🥳🤗 [image]
  • @karpathy Andrej Karpathy on x
    New open weights LLM from @MistralAI params.json: - hidden_dim / dim = 14336/4096 => 3.5X MLP expand - n_heads / n_kv_heads = 32/8 => 4X multiquery - “moe” => mixture of experts 8X top 2 👀 Likely related code: https://github.com/... Oddly absent: an over-rehearsed... [image]
  • @realgeorgehotz George Hotz on x
    Google put out a press release and a fake demo. Mistral put out a torrent.
  • @levelsio @levelsio on x
    Open source LLMs reaching GPT-4 levels way earlier than we thought is the most exciting thing now to me With open source RAGs (plugins for LLMs) we can even add stuff like GPT-4 Vision, web browsing and data analysis and reach features parity with GPT-4 and future versions This..…
  • @josephjacks_ @josephjacks_ on x
    Incredible that @MistralAI just raised $400M~ on a $2B post-money valuation in this market with a purely open source product (so far), no revenue and < 30 employees. Well done!
  • @abacaj Anton on x
    > be mistral > drop not one but two torrents (probably SOTA again) > nobody knows how to run the new model (yet) > proceed to raise another $400M same day [image]
  • @lulumeservey Lulu Cheng Meservey on x
    Google vs. Mistral: a tale of two AI launches and a case study in knowing your audience Google announced their new model two days ago. It was named Gemini, and it wafted in with a blog post, brand guidelines, a press tour, and a polished sizzle reel that later turned out to be...…
  • @jxmnop Jack Morris on x
    say what you will about mistral, tweeting exclusively download links to new models with no context is unbelievably cool [image]
  • @unsorsodicorda Andrea Panizza on x
    TL;DR: 1) sparse MoE, running 2 experts/token (we all knew that 😂) 2) marginally better than GPT-3.5 3) there's a Mixtral Instruct 4) there's also a Mistral Medium, which is probably way bigger than Mixtral, and way better than GPT 3.5 https://mistral.ai/... so much for 1/n
  • @maccaw Alex MacCaw on x
    Looks like we might see an open source GPT-4 level model much sooner than I thought. Get the Starling team fine tuning Mistral.
  • @nicoritschel Nico Ritschel on x
    So there's a 3rd @MistralAI model that outperforms Friday's mixtral (and GPT-3.5). Also hosted inference! https://mistral.ai/... [image]
  • @abacaj Anton on x
    After playing with this new mistral model for the last 24 hours, pretty sure if you go through fine tuning and rlhf you would get a > gpt-3.5 local model
  • @arthurmensch Arthur Mensch on x
    Announcing Mixtral 8x7B https://mistral.ai/... and our early developer platform https://mistral.ai/.... Very proud of the team!
  • @basedbeffjezos @basedbeffjezos on x
    OSS LLMs competing with big centralized co models is the future we want.
  • @vikhyatk Vik on x
    Mistral's stated goal for this model (according to their pitch deck) was to beat ChatGPT 3.5 by a large margin. [image]