August 6, 2025 at 11:42 AM
GPT-OSS Swallow
https://swallow-llm.github.io/gptoss-swallow.ja.html
https://swallow-llm.github.io/gptoss-swallow.ja.html
GPT-OSS Swallow
GPT-OSSの日本語能力と推論能力を強化した大規模言語モデル (20B, 120B)
swallow-llm.github.io
February 20, 2026 at 1:26 PM
GPT-OSS Swallow
https://swallow-llm.github.io/gptoss-swallow.ja.html
https://swallow-llm.github.io/gptoss-swallow.ja.html
OpenAI Returns to Open-Weight AI with New gpt-oss Models
#AI #OpenAI #OpenSource #LLM #GptOSS #OpenSourceAI #OpenWeight #GenAI
winbuzzer.com/2025/08/05/o...
#AI #OpenAI #OpenSource #LLM #GptOSS #OpenSourceAI #OpenWeight #GenAI
winbuzzer.com/2025/08/05/o...
August 5, 2025 at 6:08 PM
OpenAI Returns to Open-Weight AI with New gpt-oss Models
#AI #OpenAI #OpenSource #LLM #GptOSS #OpenSourceAI #OpenWeight #GenAI
winbuzzer.com/2025/08/05/o...
#AI #OpenAI #OpenSource #LLM #GptOSS #OpenSourceAI #OpenWeight #GenAI
winbuzzer.com/2025/08/05/o...
Last year my LLM prediction was that local models will become more and more ubiquitous
Good to see the validation of that with OpenAI GPTOSS and Simon Willson's recent blog post touching on the same pattern of local-first LLM models. Inference tech will be key.
Good to see the validation of that with OpenAI GPTOSS and Simon Willson's recent blog post touching on the same pattern of local-first LLM models. Inference tech will be key.
September 8, 2025 at 6:00 PM
Last year my LLM prediction was that local models will become more and more ubiquitous
Good to see the validation of that with OpenAI GPTOSS and Simon Willson's recent blog post touching on the same pattern of local-first LLM models. Inference tech will be key.
Good to see the validation of that with OpenAI GPTOSS and Simon Willson's recent blog post touching on the same pattern of local-first LLM models. Inference tech will be key.
日本語能力を強化したAI「GPT-OSS Swallow」と「Qwen3 Swallow」を東京科学大の研究チームが公開
https://gigazine.net/news/20260224-gptoss-qwen3-swallow-japanese-ai/
https://gigazine.net/news/20260224-gptoss-qwen3-swallow-japanese-ai/
日本語能力を強化したAI「GPT-OSS Swallow」と「Qwen3 Swallow」を東京科学大の研究チームが公開
2026年2月20日、東京科学大学情報理工学院の岡崎研究室と横田研究室、国立研究開発法人産業技術総合研究所(産総研：AIST)の研究チームが、OpenAI GPT-OSSの日本語能力と思考力を強化した推論型大規模言語モデルの「GPT-OSS Swallow」と、Alibaba Qwen3の日本語能力と思考力を強化した推論型大規模言語モデルの「Qwen3 Swallow」を発表しました。
gigazine.net
February 24, 2026 at 12:00 PM
日本語能力を強化したAI「GPT-OSS Swallow」と「Qwen3 Swallow」を東京科学大の研究チームが公開
https://gigazine.net/news/20260224-gptoss-qwen3-swallow-japanese-ai/
https://gigazine.net/news/20260224-gptoss-qwen3-swallow-japanese-ai/
Robert Langdon’s next thriller will be to solve Neuro-symbolic computing with gptOSS running on a Mac Mini.
October 28, 2025 at 10:56 AM
Robert Langdon’s next thriller will be to solve Neuro-symbolic computing with gptOSS running on a Mac Mini.
Based on the InferenceMAX blog post, it seems they are accounting for speculative decoding and MTP. InferenceMAX doesn't test gpt-oss on GB200, maybe that would have given faster speeds. It also could be shorter requests or different in/out ratio maybe?
InferenceMAX™: Open Source Inference Benchmarking
NVIDIA GB200 NVL72, AMD MI355X, Throughput Token per GPU, Latency Tok/s/user, Perf per Dollar, Tokens per Provisioned Megawatt, DeepSeek R1 670B, GPTOSS 120B, Llama3 70B
newsletter.semianalysis.com
November 8, 2025 at 4:27 PM
Based on the InferenceMAX blog post, it seems they are accounting for speculative decoding and MTP. InferenceMAX doesn't test gpt-oss on GB200, maybe that would have given faster speeds. It also could be shorter requests or different in/out ratio maybe?
GPT-OSS is the talk of the (dev) town.
This visual-heavy newsletter is a great breakdown of:
– Reasoning modes
– Message formatting
– How GPT-OSS compares to Qwen, DeepSeek, and others
👀: bit.ly/45EjKl3
And yes - you can run it locally with Docker Model Runner.
#GPTOSS #Docker #LLM #AIInfra
This visual-heavy newsletter is a great breakdown of:
– Reasoning modes
– Message formatting
– How GPT-OSS compares to Qwen, DeepSeek, and others
👀: bit.ly/45EjKl3
And yes - you can run it locally with Docker Model Runner.
#GPTOSS #Docker #LLM #AIInfra
The Illustrated GPT-OSS
OpenAI releases their first open source LLM in six years
bit.ly
August 22, 2025 at 12:10 AM
GPT-OSS is the talk of the (dev) town.
This visual-heavy newsletter is a great breakdown of:
– Reasoning modes
– Message formatting
– How GPT-OSS compares to Qwen, DeepSeek, and others
👀: bit.ly/45EjKl3
And yes - you can run it locally with Docker Model Runner.
#GPTOSS #Docker #LLM #AIInfra
This visual-heavy newsletter is a great breakdown of:
– Reasoning modes
– Message formatting
– How GPT-OSS compares to Qwen, DeepSeek, and others
👀: bit.ly/45EjKl3
And yes - you can run it locally with Docker Model Runner.
#GPTOSS #Docker #LLM #AIInfra
GPT-OSS – A C# Guide with Ollama | by Bruno Capuano.
buff.ly/IF1bWFD
#dotnet #csharp #ollama #ai #gptoss #aimodels
buff.ly/IF1bWFD
#dotnet #csharp #ollama #ai #gptoss #aimodels
GPT-OSS - A C# Guide with Ollama - .NET Blog
Run GPT-OSS locally with C# and Ollama to build fast, private, offline AI
buff.ly
August 20, 2025 at 5:00 PM
Here's their blog post on the benchmark. I've only skimmed it, but they seem to expect significant future speedups due to software improvements. newsletter.semianalysis.com/p/inferencem...
InferenceMAX™: Open Source Inference Benchmarking
NVIDIA GB200 NVL72, AMD MI355X, Throughput Token per GPU, Latency Tok/s/user, Perf per Dollar, Tokens per Provisioned Megawatt, DeepSeek R1 670B, GPTOSS 120B, Llama3 70B
newsletter.semianalysis.com
October 10, 2025 at 4:34 PM
Here's their blog post on the benchmark. I've only skimmed it, but they seem to expect significant future speedups due to software improvements. newsletter.semianalysis.com/p/inferencem...
Open models d’ #OpenAI
👉#OpenSource #GPTOSS tourne avec très peu de ressources et serait au niveau que 4o-mini
👉Des modèles de raisonnement qui peuvent être adaptés à tous les cas d’utilisation et tous les environnements d’exécution
openai.com/fr-FR/open-m...
👉#OpenSource #GPTOSS tourne avec très peu de ressources et serait au niveau que 4o-mini
👉Des modèles de raisonnement qui peuvent être adaptés à tous les cas d’utilisation et tous les environnements d’exécution
openai.com/fr-FR/open-m...
Open models d’OpenAI
Advanced open-weight reasoning models to customize for any use case and run anywhere.
openai.com
August 6, 2025 at 5:25 PM
Open models d’ #OpenAI
👉#OpenSource #GPTOSS tourne avec très peu de ressources et serait au niveau que 4o-mini
👉Des modèles de raisonnement qui peuvent être adaptés à tous les cas d’utilisation et tous les environnements d’exécution
openai.com/fr-FR/open-m...
👉#OpenSource #GPTOSS tourne avec très peu de ressources et serait au niveau que 4o-mini
👉Des modèles de raisonnement qui peuvent être adaptés à tous les cas d’utilisation et tous les environnements d’exécution
openai.com/fr-FR/open-m...
🚀 GPT-OSS is booming:
5M+ downloads, 400+ Hugging Face models in a week.
Want to run it locally, even without a GPU?
Docker's Developer Advocate Ajeet Raina shows you how using Docker Offload:
bit.ly/4fF6bVA
#Docker #GPTOSS #DockerOffload #LLM #OpenSourceAI
5M+ downloads, 400+ Hugging Face models in a week.
Want to run it locally, even without a GPU?
Docker's Developer Advocate Ajeet Raina shows you how using Docker Offload:
bit.ly/4fF6bVA
#Docker #GPTOSS #DockerOffload #LLM #OpenSourceAI
How to Run OpenAI's New GPT-OSS Models using Docker Offload
Ready to revolutionize your AI workflow? Start your GPT-OSS journey today and join the open AI revolution.
bit.ly
August 18, 2025 at 4:27 PM
🚀 GPT-OSS is booming:
5M+ downloads, 400+ Hugging Face models in a week.
Want to run it locally, even without a GPU?
Docker's Developer Advocate Ajeet Raina shows you how using Docker Offload:
bit.ly/4fF6bVA
#Docker #GPTOSS #DockerOffload #LLM #OpenSourceAI
5M+ downloads, 400+ Hugging Face models in a week.
Want to run it locally, even without a GPU?
Docker's Developer Advocate Ajeet Raina shows you how using Docker Offload:
bit.ly/4fF6bVA
#Docker #GPTOSS #DockerOffload #LLM #OpenSourceAI
🔹 New GPT-OSS test
Tried it!
Turns out my GPU is not a fan of 16GB requirements. 🤓
#GPTOSS #AI #LLM #GenerativeAI
Tried it!
Turns out my GPU is not a fan of 16GB requirements. 🤓
#GPTOSS #AI #LLM #GenerativeAI
August 11, 2025 at 8:45 AM
🔹 New GPT-OSS test
Tried it!
Turns out my GPU is not a fan of 16GB requirements. 🤓
#GPTOSS #AI #LLM #GenerativeAI
Tried it!
Turns out my GPU is not a fan of 16GB requirements. 🤓
#GPTOSS #AI #LLM #GenerativeAI
日本語性能を強化したオープンなLLM「GPT-OSS Swallow」と「Qwen3 Swallow」リリース | gihyo.jp
https://gihyo.jp/article/2026/02/gptoss-qwen3-swallow
https://gihyo.jp/article/2026/02/gptoss-qwen3-swallow
日本語性能を強化したオープンなLLM「GPT-OSS Swallow」と「Qwen3 Swallow」リリース | gihyo.jp
Swallow LLM Projectは2026年2月20日、OpenAI GPT-OSSおよびAlibaba Qwen3の日本語能力と思考力を強化した推論型言語モデル「GPT-OSS Swallow」と「Qwen3 Swallow」をリリースした。
gihyo.jp
February 24, 2026 at 8:45 PM
日本語性能を強化したオープンなLLM「GPT-OSS Swallow」と「Qwen3 Swallow」リリース | gihyo.jp
https://gihyo.jp/article/2026/02/gptoss-qwen3-swallow
https://gihyo.jp/article/2026/02/gptoss-qwen3-swallow
Microsoft Brings OpenAI’s gpt-oss Models to Windows with New Azure AI Foundry Local Platform
#AI #OnDeviceAI #OpenAI #gptoss #Microsoft #Windows #Developers #Devs #AzureFoundry #AzureFoundryLocal
winbuzzer.com/2025/08/05/m...
#AI #OnDeviceAI #OpenAI #gptoss #Microsoft #Windows #Developers #Devs #AzureFoundry #AzureFoundryLocal
winbuzzer.com/2025/08/05/m...
August 5, 2025 at 9:54 PM
Microsoft Brings OpenAI’s gpt-oss Models to Windows with New Azure AI Foundry Local Platform
#AI #OnDeviceAI #OpenAI #gptoss #Microsoft #Windows #Developers #Devs #AzureFoundry #AzureFoundryLocal
winbuzzer.com/2025/08/05/m...
#AI #OnDeviceAI #OpenAI #gptoss #Microsoft #Windows #Developers #Devs #AzureFoundry #AzureFoundryLocal
winbuzzer.com/2025/08/05/m...
Effectivement, d'autant que les abonnements d'Anthropic sont assez chers. Un ami fait tourner un gptoss-120B en local sur CPU, et a configuré claude code pour tourner dessus, il en est content.
Par contre il faut plus de 64 Go de RAM rien que pour le modèle ^^
Par contre il faut plus de 64 Go de RAM rien que pour le modèle ^^
January 4, 2026 at 5:01 PM
Effectivement, d'autant que les abonnements d'Anthropic sont assez chers. Un ami fait tourner un gptoss-120B en local sur CPU, et a configuré claude code pour tourner dessus, il en est content.
Par contre il faut plus de 64 Go de RAM rien que pour le modèle ^^
Par contre il faut plus de 64 Go de RAM rien que pour le modèle ^^
OpenAI's move empowers developers worldwide—from indie creators to enterprises—to build and innovate with powerful, safe AI on their own terms.
#OpenAI #GPTOss #AIInnovation #OpenSourceAI #techgalena #MachineLearning #TechNews #ArtificialIntelligence #FutureOfAI
#OpenAI #GPTOss #AIInnovation #OpenSourceAI #techgalena #MachineLearning #TechNews #ArtificialIntelligence #FutureOfAI
August 7, 2025 at 7:15 AM
OpenAI's move empowers developers worldwide—from indie creators to enterprises—to build and innovate with powerful, safe AI on their own terms.
#OpenAI #GPTOss #AIInnovation #OpenSourceAI #techgalena #MachineLearning #TechNews #ArtificialIntelligence #FutureOfAI
#OpenAI #GPTOss #AIInnovation #OpenSourceAI #techgalena #MachineLearning #TechNews #ArtificialIntelligence #FutureOfAI
✨ OpenAI & You.com: Neue KI-Ära 2025?
▶️ Open-Weight Modelle starten
▶️ Live-Websuche integriert
▶️ Effizient & faktenbasiert
#ai #ki #artificialintelligence #openai #youcom #gptoss #opensourceai #tech2025
⚡ SAVE IT! SHARE IT! READ IT! 🚀
kinews24.de/openai-gpt-o...
▶️ Open-Weight Modelle starten
▶️ Live-Websuche integriert
▶️ Effizient & faktenbasiert
#ai #ki #artificialintelligence #openai #youcom #gptoss #opensourceai #tech2025
⚡ SAVE IT! SHARE IT! READ IT! 🚀
kinews24.de/openai-gpt-o...
September 9, 2025 at 1:51 PM
✨ OpenAI & You.com: Neue KI-Ära 2025?
▶️ Open-Weight Modelle starten
▶️ Live-Websuche integriert
▶️ Effizient & faktenbasiert
#ai #ki #artificialintelligence #openai #youcom #gptoss #opensourceai #tech2025
⚡ SAVE IT! SHARE IT! READ IT! 🚀
kinews24.de/openai-gpt-o...
▶️ Open-Weight Modelle starten
▶️ Live-Websuche integriert
▶️ Effizient & faktenbasiert
#ai #ki #artificialintelligence #openai #youcom #gptoss #opensourceai #tech2025
⚡ SAVE IT! SHARE IT! READ IT! 🚀
kinews24.de/openai-gpt-o...
🔥 Want to try the new OpenAI gpt-oss model in C#?
🚀 Here’s a ready-to-use Ollama sample:
- aka.ms/genainet
- openai.com/index/introd...
Runs gpt-oss:20b locally + Microsoft.Extensions.AI
#dotnet #CSharp #AI #OpenAI #Ollama #gptoss
🚀 Here’s a ready-to-use Ollama sample:
- aka.ms/genainet
- openai.com/index/introd...
Runs gpt-oss:20b locally + Microsoft.Extensions.AI
#dotnet #CSharp #AI #OpenAI #Ollama #gptoss
GitHub - microsoft/Generative-AI-for-beginners-dotnet: Five lessons, learn how to really apply AI to your .NET Applications
Five lessons, learn how to really apply AI to your .NET Applications - microsoft/Generative-AI-for-beginners-dotnet
aka.ms
August 5, 2025 at 8:32 PM
🔥 Want to try the new OpenAI gpt-oss model in C#?
🚀 Here’s a ready-to-use Ollama sample:
- aka.ms/genainet
- openai.com/index/introd...
Runs gpt-oss:20b locally + Microsoft.Extensions.AI
#dotnet #CSharp #AI #OpenAI #Ollama #gptoss
🚀 Here’s a ready-to-use Ollama sample:
- aka.ms/genainet
- openai.com/index/introd...
Runs gpt-oss:20b locally + Microsoft.Extensions.AI
#dotnet #CSharp #AI #OpenAI #Ollama #gptoss
── GPT-OSS Swallow 20Bの日本語タスクの平均スコアは0.606で、総パラメータ数が20B以下のオープンなLLMの中で最高性能を達成しました。基にしたgpt-oss-20bと比較すると、ほぼ全てのタスクで性能向上が得られました
swallow-llm.github.io/gptoss-swall...
すごそう。トレードオフなしで日本語能力以外も含めて完全強化した例はかなり珍しいのでは。
swallow-llm.github.io/gptoss-swall...
すごそう。トレードオフなしで日本語能力以外も含めて完全強化した例はかなり珍しいのでは。
GPT-OSS Swallow
GPT-OSSの日本語能力と推論能力を強化した大規模言語モデル (20B, 120B)
swallow-llm.github.io
February 24, 2026 at 2:29 PM
── GPT-OSS Swallow 20Bの日本語タスクの平均スコアは0.606で、総パラメータ数が20B以下のオープンなLLMの中で最高性能を達成しました。基にしたgpt-oss-20bと比較すると、ほぼ全てのタスクで性能向上が得られました
swallow-llm.github.io/gptoss-swall...
すごそう。トレードオフなしで日本語能力以外も含めて完全強化した例はかなり珍しいのでは。
swallow-llm.github.io/gptoss-swall...
すごそう。トレードオフなしで日本語能力以外も含めて完全強化した例はかなり珍しいのでは。
The neat part is how controlled it is:
Homelab questions stay inside the RAG world, but general topics only switch to model-knowledge when I explicitly allow it through routing at the RAG layer.
Feels like an assistant that actually stays in its lane.
#EdgeAI #LocalAI #GPTOSS #TechStack #homelab
Homelab questions stay inside the RAG world, but general topics only switch to model-knowledge when I explicitly allow it through routing at the RAG layer.
Feels like an assistant that actually stays in its lane.
#EdgeAI #LocalAI #GPTOSS #TechStack #homelab
November 23, 2025 at 4:42 AM
The neat part is how controlled it is:
Homelab questions stay inside the RAG world, but general topics only switch to model-knowledge when I explicitly allow it through routing at the RAG layer.
Feels like an assistant that actually stays in its lane.
#EdgeAI #LocalAI #GPTOSS #TechStack #homelab
Homelab questions stay inside the RAG world, but general topics only switch to model-knowledge when I explicitly allow it through routing at the RAG layer.
Feels like an assistant that actually stays in its lane.
#EdgeAI #LocalAI #GPTOSS #TechStack #homelab