Rohit Tiwari (@_rohit_tiwari_) 's Twitter Profile
Rohit Tiwari

@_rohit_tiwari_

Passionate about building intelligent machine and deep learning systems.

ID: 1263903700998709249

calendar_today22-05-2020 18:45:34

40 Tweet

11 Followers

287 Following

Rohit Tiwari (@_rohit_tiwari_) 's Twitter Profile Photo

GPT-OSS is still a big deal Even GPT-5 is here, why? This guide will save you hours. OpenAI has recently released two open-weight mixture of experts (MoE) reasoning models.  ↳ gpt-oss-20b  ↳ gpt-oss-120b Reasoning 🟩 🟩 🟩 🟩 Coding 🟩 🟩 🟩 🟩 🟩 Math

GPT-OSS is still a big deal
Even GPT-5 is here, why?
This guide will save you hours.

OpenAI has recently released two open-weight mixture of experts (MoE) reasoning models.
 ↳ gpt-oss-20b
 ↳ gpt-oss-120b

Reasoning 
🟩 🟩 🟩 🟩 

Coding          
🟩 🟩 🟩 🟩 🟩

Math
Rohit Tiwari (@_rohit_tiwari_) 's Twitter Profile Photo

GPT-5 reasoning on ARC-AGI 2 and LMArena benchmarks. The results are: • GPT-5 reasoning: 70% • GPT-5 base: 25% • Grok 4: 17% • o3-pro: 4% • Gemini 2.5 Pro: 3% It is the best reasoning model as of today.

GPT-5 reasoning on ARC-AGI 2 and LMArena benchmarks.

The results are:

• GPT-5 reasoning: 70%
• GPT-5 base: 25%
• Grok 4: 17%
• o3-pro: 4%
• Gemini 2.5 Pro: 3%

It is the best reasoning model as of today.
Rohit Tiwari (@_rohit_tiwari_) 's Twitter Profile Photo

GPT-5 reasoning on ARC-AGI 2 and LMArena benchmarks. The results are: • GPT-5 reasoning: 70% • GPT-5 base: 25% • Grok 4: 17% • o3-pro: 4% • Gemini 2.5 Pro: 3% It is the best reasoning model as of today.

GPT-5 reasoning on ARC-AGI 2 and LMArena benchmarks.

The results are:

• GPT-5 reasoning: 70%
• GPT-5 base: 25%
• Grok 4: 17%
• o3-pro: 4%
• Gemini 2.5 Pro: 3%

It is the best reasoning model as of today.
Rohit Tiwari (@_rohit_tiwari_) 's Twitter Profile Photo

Marcus Cvjeticanin The gpt-5 release suggests that we are nearing the saturation point for LLMs. Further major progress will likely come from breakthroughs in underlying architecture. Overall, gpt-5 is a decent upgrade. Just few days back, OpenAI also released the open weight gpt-oss models (20b