Trending topics
#
Bonk Eco continues to show strength amid $USELESS rally
#
Pump.fun to raise $1B token sale, traders speculating on airdrop
#
Boop.Fun leading the way with a new launchpad on Solana.
Understand the new DeepSeek model DeepSeek-Math-V2 at a glance!
DeepSeek has just released DeepSeek-Math-V2!
That's right, this is a domain-specific model aimed at mathematical reasoning, and this time it has been post-trained/fine-tuned based on DeepSeek-V3.2, meaning DeepSeek has completely switched to the new architecture.
In terms of performance, it scored 83.3% in IMO 2025 (International Mathematical Olympiad), reaching the gold medal level. ProofBench-Basic scored a staggering 99%, crushing all other models. ProofBench-Advanced came in just behind Gemini Deep Think (which is a platform with an Agent, not just a pure model).
By the way, this year's IMO 2025 first place also belongs to our country. Then I averaged the score of 83.3% from DeepSeek-Math-V2 across each question, which should be 210 points (out of a total of 252 points, with 83.3% accuracy), placing it between the United States (216 points) and South Korea (203 points), in third place.

Top
Ranking
Favorites

