- Bay Area Times
- Posts
- OpenAI releases first open-weight models since GPT-2 in major shift
OpenAI releases first open-weight models since GPT-2 in major shift


Top stories today:
- OpenAI releases first open-weight models since GPT-2 in major shift
- OpenAI said to be in talks for share sale at $500B valuation
- Anthropic launches Claude Opus 4.1 with broad improvements
- Google DeepMind unveils Genie 3 āworldā model to generate 3D worlds
- Pfizer Q2 revenue +10% YoY to $14.7B, lifts 2025 outlook on cost cuts
0. Data and calendar

All values as of 6 AM ET / 3 AM PT, other than S&P500 and NASDAQ close (4 PM ET / 1 PM PT).

All times are ET.
Listen to our AI-generated podcast summarizing todayās newsletter (beware of hallucinations):
1. OpenAI releases first open-weight models since GPT-2 in major shift
gpt-oss-120b and gpt-oss-20b are OpenAIās two open-weight models, scoring 58 and 48, respectively, on the Artificial Analysis Intelligence Index:
Both models are Apache 2.0 licensed and available through Hugging Face.
gpt-oss-20b can even run locally on a consumer laptop with >16GB RAM, and gpt-oss-120b on higher-end machines.
The models use āchain-of-thought reasoningā approach, which was earlier deployed in the o-series.
2019: GPT-2 was launched as OpenAIās last open-weight model.
OpenAIās gpt-oss-120b is leading U.S. open-weight models, while Alibabaās Qwen3 235B is topping globally
The 120B model is significantly smaller than competing open-weight models, with 116.8B total parameters and 5.1B active parameters.
gpt-oss-120B is also the most intelligence model that can fit on a single H100 GPU in its native precision
At API providers, the median price of the 120B is $0.15/M input and $0.69/M output tokens, ~10Ć cheaper than o4-mini ($1.1/$4.4)
$0.08/$0.35 is the 20B modelās median price.
2. OpenAI said to be in talks for share sale at $500B valuation

Existing investors like Thrive Capital have reportedly sought to buy OpenAI employee shares.
A secondary sale could help OpenAI retain staff amid talent poaching by Meta and others.
Struggling to find high-quality ML engineers? Athyna connects you with top Machine Learning talent from LATAM. AI-matched and over 50% more affordable than Silicon Valley.
Matched for quality, speed & fit.
Based in Latin America for optimal overlap.
Build smarter, save bigger.
*Disclaimer: We have equity in Athyna.
4. Anthropic launches Claude Opus 4.1 AI model with broad improvements at same cost
Opus 4.1 has debuted as an upgrade to Opus 4, with better coding performance: