#devstral — Public Fediverse posts
Live and recent posts from across the Fediverse tagged #devstral, aggregated by home.social.
-
Before working for 2 years on the Apache APISIX API gateway, I was mainly oblivious to API gateways. It’s only by working with them that I understood their value. Decoupling the client and the server unlocks a lot of options: moving authentication to the API Gateway, securing APIs, deduplicating API requests, etc.
In this post, I want to describe how the same pattern applies to #AI.
#AIGateways work in a similar way.
-
Before working for 2 years on the Apache APISIX API gateway, I was mainly oblivious to API gateways. It’s only by working with them that I understood their value. Decoupling the client and the server unlocks a lot of options: moving authentication to the API Gateway, securing APIs, deduplicating API requests, etc.
In this post, I want to describe how the same pattern applies to #AI.
#AIGateways work in a similar way.
-
RE: https://norden.social/@maik/116328948179344820
Artikel wie schon von Arstechnica. Die interessantere Frage ist wie lange es dauern wird bist auf dieser Basis Angriffe auf / mit / durch Claude Code gefahren werden. Und was für Effekte daraus zu sehen sein werden.
Abgesehen davon: Ich bin hier um mit Ihnen über das Wort Go...äh, OpenCode zu reden: https://opencode.ai/
Gleiches LLM coding, allerdings was Modelle angeht offen und inhärent gleich FOSS als Codebasis. Funktioniert auch mit Mistral, ist deutlich weiter entwickelt als Mistrals VibeCLI.
Kritik und Vorteile alle wie gehabt, mehr Auswahl und mehr Offenheit.#ai #claude #claudecode #opencode #mistral #mistralvibe #codestral #devstral2 #devstral #llm #vibecoding
-
RE: https://norden.social/@maik/116328948179344820
Artikel wie schon von Arstechnica. Die interessantere Frage ist wie lange es dauern wird bist auf dieser Basis Angriffe auf / mit / durch Claude Code gefahren werden. Und was für Effekte daraus zu sehen sein werden.
Abgesehen davon: Ich bin hier um mit Ihnen über das Wort Go...äh, OpenCode zu reden: https://opencode.ai/
Gleiches LLM coding, allerdings was Modelle angeht offen und inhärent gleich FOSS als Codebasis. Funktioniert auch mit Mistral, ist deutlich weiter entwickelt als Mistrals VibeCLI.
Kritik und Vorteile alle wie gehabt, mehr Auswahl und mehr Offenheit.#ai #claude #claudecode #opencode #mistral #mistralvibe #codestral #devstral2 #devstral #llm #vibecoding
-
@Bit @plexus Well, after a few days with it, I'm going back to using Claude Code with Ollama. Opencode was just too brittle on my setup. Who knows, I might try again later.
Here's the script I'm using to call #devstral in #ollama from #claudecode
https://gist.github.com/rubix1138/be281b66971b02563fd4be5bd785b9dd -
@Bit @plexus Well, after a few days with it, I'm going back to using Claude Code with Ollama. Opencode was just too brittle on my setup. Who knows, I might try again later.
Here's the script I'm using to call #devstral in #ollama from #claudecode
https://gist.github.com/rubix1138/be281b66971b02563fd4be5bd785b9dd -
Playing with #OpenHands and #Devstral from #MistralAI. I already hit a RateLimitError 429 ☹️ I'm also missing the time I can resume my work. And yes, I have activated the tokens compression settings.
It's not today that I'll be able to evaluate whether it's better than Claude Code or not.
-
Playing with #OpenHands and #Devstral from #MistralAI. I already hit a RateLimitError 429 ☹️ I'm also missing the time I can resume my work. And yes, I have activated the tokens compression settings.
It's not today that I'll be able to evaluate whether it's better than Claude Code or not.
-
-
The internet(tm) claims that DevStral2 would be almost on par with Claude Sonnet 4.5.
Also, open weights model with 120B, i.e. reasonable on current om-prem hardware a la AMDs 395+.
Prices are also pretty ok if hosted, also: EU hostable.Did anyone compare that thing to the above Claude model, ideally using Claude Code as control system?
I'd also be interested in on-prem reports, ideally also with one of the AMD 395+ UMA boxes.#llm #claudecode #claude_sonnet_45 #sonnet45 #devstral #devstral2 #mistral #aicoding #uma #amd #amdstrixhalo
-
The internet(tm) claims that DevStral2 would be almost on par with Claude Sonnet 4.5.
Also, open weights model with 120B, i.e. reasonable on current om-prem hardware a la AMDs 395+.
Prices are also pretty ok if hosted, also: EU hostable.Did anyone compare that thing to the above Claude model, ideally using Claude Code as control system?
I'd also be interested in on-prem reports, ideally also with one of the AMD 395+ UMA boxes.#llm #claudecode #claude_sonnet_45 #sonnet45 #devstral #devstral2 #mistral #aicoding #uma #amd #amdstrixhalo
-
Mistral Vibe 2.0 bringt KI-Agenten in das Terminal. Anstatt auf IDE-Plugins zu setzen, integriert sich das Tool direkt in die Shell und Dateisysteme. Spezialisierte Sub-Agenten werden über Slash-Commands gesteuert. Ein neues Klärungssystem verhindert Halluzinationen durch Auswahlmenüs bei Unklarheiten. Zugriff auf das volle Tool erfordert kostenpflichtige Pläne. #MistralVibe #MistralAI #Devstral
https://www.all-ai.de/news/news26top/mistral-vibe-2 -
Got Mistral's vibe-cli up and running towards a local Ollama instance with devstral-2-small. Let's see if it can understand my retro game coding level editor (written in Java).
-
Got Mistral's vibe-cli up and running towards a local Ollama instance with devstral-2-small. Let's see if it can understand my retro game coding level editor (written in Java).
-
🔧 #OpenRouter launches Response Healing: automatically fixes malformed #JSON responses from #LLMs before they reach your app #AI #API #devops
📊 Results from 1.6M+ requests analyzed at inference time:
• #Gemini 2.0 Flash: 80% defect reduction (99.61% → 99.92%)
• #Qwen3 235B: 99.8% defect reduction (88% → 99.98%)
• #Deepseek Chat v3.1: 85% defect reduction (82.5% → 97.4%)
• #Devstral 2512: 99.6% defect reduction (96.6% → 99.99%)🧵 👇
-
🔧 #OpenRouter launches Response Healing: automatically fixes malformed #JSON responses from #LLMs before they reach your app #AI #API #devops
📊 Results from 1.6M+ requests analyzed at inference time:
• #Gemini 2.0 Flash: 80% defect reduction (99.61% → 99.92%)
• #Qwen3 235B: 99.8% defect reduction (88% → 99.98%)
• #Deepseek Chat v3.1: 85% defect reduction (82.5% → 97.4%)
• #Devstral 2512: 99.6% defect reduction (96.6% → 99.99%)🧵 👇
-
Mistral released new coding models:
* Devstral 2 (123 B)
* Devstral Small 2 (24 B)https://mistral.ai/news/devstral-2-vibe-cli
Plus, Mistral Vibe CLI, a CLI based coding agent:
https://github.com/mistralai/mistral-vibeI didn't even have a chance yet to try out Mistral 3 Large. According to their benchmark, Devstral 2 is competing with the best open weight models (DeepSeek V3.2). Not quite at the level of the best closed-sources ones, but more cost-efficient:
> Devstral 2 is currently offered free via our API. After the free period, the API pricing will be $0.40/$2.00 per million tokens (input/output) for Devstral 2 and $0.10/$0.30 for Devstral Small 2.
Devstral Small 2 can be run locally on a 24 GB GPU.
-
Mistral released new coding models:
* Devstral 2 (123 B)
* Devstral Small 2 (24 B)https://mistral.ai/news/devstral-2-vibe-cli
Plus, Mistral Vibe CLI, a CLI based coding agent:
https://github.com/mistralai/mistral-vibeI didn't even have a chance yet to try out Mistral 3 Large. According to their benchmark, Devstral 2 is competing with the best open weight models (DeepSeek V3.2). Not quite at the level of the best closed-sources ones, but more cost-efficient:
> Devstral 2 is currently offered free via our API. After the free period, the API pricing will be $0.40/$2.00 per million tokens (input/output) for Devstral 2 and $0.10/$0.30 for Devstral Small 2.
Devstral Small 2 can be run locally on a 24 GB GPU.
-
La #start-up #française #Mistral se lance dans la course au #vibe-coding et publie #Devstral 2 sous une #licence #MIT modifiée? une nouvelle génération de son modèle d' #IA #AI conçu pour le #codage #dev #tech #FrenchTech #frenchtech #france intelligence-artificielle.developpez.com/actu/378346/...
La start-up fran�aise Mistral ... -
Нейро-дайджест: ключевые события мира AI за 2-ю неделю декабря 2025
Привет! Это новый выпуск «Нейро-дайджеста» — коротких и полезных обзоров ключевых событий в мире искусственного интеллекта и технологий. Неделя выдалась насыщенной: Mistral выкатили Devstral 2 , Photoshop, Express и Acrobat встроили прямо в ChatGPT , а исследования показывают, что более 80% вайб-кода могут содержать уязвимости. Криштиану Роналду инвестирует в Perplexity, а Сэм Альтман считает, что сейчас невозможно воспитывать детей без ИИ. Всё самое важное — в одном месте. Поехали! Читать дайджест →
https://habr.com/ru/companies/timeweb/articles/975520/
#новости #ии #дайджест #нейросети #devstral #openrouter #timeweb_дайджест #искусственный_интеллект #perplexity #сэм_альтман
-
✨ Behold, a groundbreaking update from the Future where #Mistral releases #Devstral 2 with a whopping 72.2% of #SWE-Bench verification, because, apparently, real-world utility is soooo last century. 🤖 Meanwhile, the Vibe CLI promises to take your coding monotony to new heights of agentic glory, because who doesn't want their command line to have vibes? 📟🚀
https://mistral.ai/news/devstral-2-vibe-cli #VibeCLI #codingInnovation #HackerNews #ngated -
✨ Behold, a groundbreaking update from the Future where #Mistral releases #Devstral 2 with a whopping 72.2% of #SWE-Bench verification, because, apparently, real-world utility is soooo last century. 🤖 Meanwhile, the Vibe CLI promises to take your coding monotony to new heights of agentic glory, because who doesn't want their command line to have vibes? 📟🚀
https://mistral.ai/news/devstral-2-vibe-cli #VibeCLI #codingInnovation #HackerNews #ngated -
Mistral Releases Devstral 2 (72.2% SWE-Bench Verified) and Vibe CLI
-
Mistral Releases Devstral 2 (72.2% SWE-Bench Verified) and Vibe CLI
-
https://winbuzzer.com/2025/12/09/ai-coding-mistral-ai-launches-devstral-2-model-and-vibe-cli-xcxwbn/
AI Coding: Mistral AI Launches Devstral 2 Model and Vibe CLI
#AI #AICoding #MistralAI #Devstral #Devstral2 #OpenSourceAI #AgenticAI #VibeCoding #LLM #GenAI#EnterpriseAI
-
https://winbuzzer.com/2025/12/09/ai-coding-mistral-ai-launches-devstral-2-model-and-vibe-cli-xcxwbn/
AI Coding: Mistral AI Launches Devstral 2 Model and Vibe CLI
#AI #AICoding #MistralAI #Devstral #Devstral2 #OpenSourceAI #AgenticAI #VibeCoding #LLM #GenAI#EnterpriseAI
-
@ankoeln @muellertadzio Ein mglw. relevanter LLM-Skill mag das selber-Hosten auf eigener Hardware sein. Geht auch für Schönschreibtools.
Dabei helfe ich im Sinne der von Herrn Matzat geforderten "Auseinandersetzung mit der Technologie" gerne, denn mit dem Punkt [wir sollten] "zugewandt, freundlich, hilfsbereit und geduldig agieren" hat er natürlich Recht.
-
@ankoeln @muellertadzio Ein mglw. relevanter LLM-Skill mag das selber-Hosten auf eigener Hardware sein. Geht auch für Schönschreibtools.
Dabei helfe ich im Sinne der von Herrn Matzat geforderten "Auseinandersetzung mit der Technologie" gerne, denn mit dem Punkt [wir sollten] "zugewandt, freundlich, hilfsbereit und geduldig agieren" hat er natürlich Recht.
-
Mistral AI: Europe’s Bold Answer to Big Tech 🇫🇷
Launched in 2023, Mistral AI is shaking up the AI world with open-source models, local trust, and rapid innovation. Its chatbot Le Chat hit 1M+ downloads after President Macron's endorsement, and its model zoo (like Devstral, Magistral, and Voxtral) is powering everything from code to speech.
#MistralAI #LeChat #OpenSourceAI #FrenchTech #AIForAll #Devstral #MistralCompute #EuropeanAI #TECHi
Read Full Article Here :- https://www.techi.com/mistral-ai-french-open-source-leader/
-
Mistral AI: Europe’s Bold Answer to Big Tech 🇫🇷
Launched in 2023, Mistral AI is shaking up the AI world with open-source models, local trust, and rapid innovation. Its chatbot Le Chat hit 1M+ downloads after President Macron's endorsement, and its model zoo (like Devstral, Magistral, and Voxtral) is powering everything from code to speech.
#MistralAI #LeChat #OpenSourceAI #FrenchTech #AIForAll #Devstral #MistralCompute #EuropeanAI #TECHi
Read Full Article Here :- https://www.techi.com/mistral-ai-french-open-source-leader/
-
> While there are many #opensource clones of #ClaudeCode out there, if your goal is to learn this technology on a deeper level, I feel there are still significant benefits to building your own agent, and working closely with the models.
Local LLM Agents - Do they work?
https://hexeditreality.com/posts/local-llm-agents/
#AI #LLM #claude #ollama #devstral #programming #softwareDevelopment #aiagent #agenticAI #agents
-
> While there are many #opensource clones of #ClaudeCode out there, if your goal is to learn this technology on a deeper level, I feel there are still significant benefits to building your own agent, and working closely with the models.
Local LLM Agents - Do they work?
https://hexeditreality.com/posts/local-llm-agents/
#AI #LLM #claude #ollama #devstral #programming #softwareDevelopment #aiagent #agenticAI #agents
-
Someone just sold his ROG FLow Z13 ($2,500) laptop for a MacBook M4 Pro in less than a day. Figured out recently.
Why? Fed up with ROCm support. "Fuck this shit". Had a meltdown morning.
Hope devs learn the lesson: software first, promises later.
#Hardware #PC #PCHardware #AI #ML #LLM #LargeLanguageModels #Qwen #Devstral #Mistral #MistralAI #ROCm #AMD #Radeon #Ryzen #RyzenAI
-
Someone just sold his ROG FLow Z13 ($2,500) laptop for a MacBook M4 Pro in less than a day. Figured out recently.
Why? Fed up with ROCm support. "Fuck this shit". Had a meltdown morning.
Hope devs learn the lesson: software first, promises later.
#Hardware #PC #PCHardware #AI #ML #LLM #LargeLanguageModels #Qwen #Devstral #Mistral #MistralAI #ROCm #AMD #Radeon #Ryzen #RyzenAI
-
#Devstral, introducing the best #OpenSource model for #codingagents.
-
#Devstral, introducing the best #OpenSource model for #codingagents.
-
Cobbled together an #ExoLabs cluster to fuck around with #devstral a bit, since it's kinda too big for my M3 Max daily driver. While in the process of bringing up nodes the model hit a bug in the #MLX #Python module that deals with inference model sharding related to passing around MLX vs Numpy data structures.
For shits and giggles and also not being a top-tier #Numpy data structure debugging guy I asked Devstral to look at the bug and figure out a fix. After one wrong turn it came up with a fix which I applied to the other nodes and now it's happily sharding the bigger Devstral models. Not sure about vibe coding as a social contagion but from a “How close are we to #Skynet”-perspective I think we're cooked, chat.
Anyway enjoy your Memorial Day weekend 🎉
Figure 1. A very heterogeneous Exo cluster.
-
Cobbled together an #ExoLabs cluster to fuck around with #devstral a bit, since it's kinda too big for my M3 Max daily driver. While in the process of bringing up nodes the model hit a bug in the #MLX #Python module that deals with inference model sharding related to passing around MLX vs Numpy data structures.
For shits and giggles and also not being a top-tier #Numpy data structure debugging guy I asked Devstral to look at the bug and figure out a fix. After one wrong turn it came up with a fix which I applied to the other nodes and now it's happily sharding the bigger Devstral models. Not sure about vibe coding as a social contagion but from a “How close are we to #Skynet”-perspective I think we're cooked, chat.
Anyway enjoy your Memorial Day weekend 🎉
Figure 1. A very heterogeneous Exo cluster.
-
📢 Don't overlook this in the wave of releases! #MistralAI has a new coding LLM: it's #Devstral, an open model perfect for on-prem, private and local deployments 🐈
📰 Have a look at the announcement: https://mistral.ai/news/devstral
-
📢 Don't overlook this in the wave of releases! #MistralAI has a new coding LLM: it's #Devstral, an open model perfect for on-prem, private and local deployments 🐈
📰 Have a look at the announcement: https://mistral.ai/news/devstral
-
Mistral AI hat gemeinsam mit AllHandsAI sein eigenes Agentic LLM für Software-Engineering-Aufgaben veröffentlicht - #Devstral.
Der Clou ist, dass es auf dem lokalen Rechner ausgeführt werden kann und dennoch, laut eigenen Benchmarks, die beste Performance aller Open-Source-Modelle erreicht.
Damit bleiben die Daten in deiner Hand.
-
Mistral AI hat gemeinsam mit AllHandsAI sein eigenes Agentic LLM für Software-Engineering-Aufgaben veröffentlicht - #Devstral.
Der Clou ist, dass es auf dem lokalen Rechner ausgeführt werden kann und dennoch, laut eigenen Benchmarks, die beste Performance aller Open-Source-Modelle erreicht.
Damit bleiben die Daten in deiner Hand.
-
#Devstral: New #opensource Model for Coding Agents by #MistralAI & #AllHandsAI 🧠
• 🏆 #Devstral achieves 46.8% on #SWEBench Verified, outperforming previous #opensource models by over 6% points and surpassing #GPT4 mini by 20%
-
#Devstral: New #opensource Model for Coding Agents by #MistralAI & #AllHandsAI 🧠
• 🏆 #Devstral achieves 46.8% on #SWEBench Verified, outperforming previous #opensource models by over 6% points and surpassing #GPT4 mini by 20%
-
🥳🎉 Behold, the #AI messiah 'Devstral' has arrived, promising to revolutionize #coding by finally making your job obsolete! 🚀🔧 Because nothing screams "cutting-edge innovation" like regurgitating code snippets under the guise of being more "agentic" than your average bear. 🙄🤖
https://mistral.ai/news/devstral #Revolution #Innovation #Devstral #Obsolete #HackerNews #ngated
