Brought to you by:
Top stories today:
- Gemini 3 Deep Think now world’s top performing model
- MiniMax releases M2.5 with cut-rate AI pricing
- Anthropic increases funding from $20B to $30B at $350B valuation
- OpenAI debuts GPT-5.3-Codex Spark as first non-NVIDIA model
- Apple stock -5% after FTC probes Apple News, Siri delays
- U.S. existing home sales -8.4% MoM to 3.91M in Jan.
0. Data and calendar
Listen to our AI-generated podcast summarizing today’s newsletter (beware of hallucinations):
1. Gemini 3 Deep Think now world’s top performing model
The Gemini 3 Deep Think, per Google, significantly beats Anthropic’s Claude Opus 4.6 and OpenAI’s GPT-5.2:
Available in the Gemini app for Ultra subscribers.
Available via the Gemini API to select researchers, engineers, and enterprises.
Developed in close partnership with scientists and researchers to tackle tough research challenges, Google said.
Google stock remained unchanged following the announcement.
2. MiniMax releases M2.5: open, comparable with Opus 4.6, but 9x cheaper
MiniMax also debuted M2.5-Lightning, offering steady throughput of 100 tokens/second, versus 50 tokens/second for standard M2.5.
Pricing: $0.30 / $2.40 per 1M input/output tokens for M2.5-Lightning and $0.30 / $1.20 for M2.5, vs. Claude Opus 4.6’s $5 / $25 and $1.75 / $14.
9x cheaper when using a 3:1 input:output token ratio, comparing to Opus 4.6.
MiniMax has rolled out three models since Oct., claiming a comparable SWE-Bench to Opus 4.6
Connecting LLMs to production systems means handling rate limits, context windows, prompt versioning, and error states most teams underestimate. You need engineers who understand the full integration stack and ship features that hold up under load.
Python and JavaScript expertise
LLM and API integration at scale
Production-tested implementations
This is the kind of talent you get with Athyna Intelligence—vetted LATAM engineers working in U.S.-aligned time zones.
*Disclaimer: We have equity in Athyna.
4. Anthropic increases funding from $20B to $30B at $380B post-money valuation
