SK Telecom A.X K2 Reaches IMO Gold-Medal Level, Ties Inkling at AIME Top
SK Telecom said its A.X K2 AI model scored 29/42 on the 2026 International Mathematical Olympiad, matching the gold-medal threshold, and tied for first on the Hugging Face MathArena AIME 2026 leaderboard with 97.1% accuracy. The company also cited prior IMO 2025 results (35/42) and reported a 9.3 BrowseComp score for web-research tasks.
How this was made
The 30-second read
Why it matters
For traders, the key question is whether these benchmark claims translate into near-term commercial traction (enterprise contracts, partnerships, or revenue guidance). The article does not provide such disclosures.
Market read
Benchmark performance may influence sentiment around reasoning-capable open-weight AI, but the article lacks financial or regulatory catalysts.
What to watch
The article notes weak BrowseComp agentic performance (9.3, last in its comparison set), which could limit adoption for web-research or agentic workflows despite strong math reasoning.
Background
SK Telecom claims Olympic-level mathematical reasoning from its A.X K2 model, evaluated on IMO 2026 and MathArena AIME 2026 leaderboards.
Ticker impact
Article says SK Telecom reported A.X K2 scored 29/42 on IMO 2026 and tied for first on MathArena AIME 2026, signaling structured-reasoning accuracy.
Near-term impact is likely limited unless accompanied by commercial contracts, guidance, or funding disclosures.
The piece is primarily performance benchmarking and methodology discussion, with no disclosed revenue, deal, or corporate action tied to SK Telecom’s stock.
Market effects
May reinforce investor focus on open-weight, structured-reasoning benchmarks (IMO/AIME) as a proxy for enterprise reliability.
Could modestly boost sentiment toward Korean AI/telecom AI initiatives if traders treat the results as credible capability signals.
Adds to the global race for reasoning-capable models, but without a disclosed commercial rollout or regulatory decision it is unlikely to reprice the whole sector.
Counterpoint
Independent vs officially graded IMO evaluation tier may reduce comparability, and leaderboard accuracy does not guarantee real-world tool-use or long-horizon reliability.
Key entities
- companySK Telecom
Reported A.X K2 benchmark results on IMO 2026 and MathArena AIME 2026, plus a limitation on BrowseComp agentic performance.
- AI modelA.X K2
Open-weight model discussed as achieving gold-medal threshold on IMO 2026 and tying for first on MathArena AIME 2026.
- AI modelInkling
Thinking Machines Lab model tied for first on the MathArena AIME 2026 leaderboard at 97.1% accuracy.
