AI Model Comparison
Side-by-side comparison of pricing, context window, benchmarks, capabilities, and license. Every figure comes from the developer's own pages, last checked on ; each model's page lists its sources. A figure we could not source shows as a dash.
Model A — LG AI Research
K-EXAONE 2.0 is a multilingual model from LG AI Research, released on July 31, 2026 and upcycled from K-EXAONE to more than three times its size, focused on reasoning, agentic workflows and long-context work. It covers ten languages, including Korean, English and Japanese.
Model B — Tencent
Hy3 is Tencent Hy's hybrid fast-and-slow-thinking Mixture-of-Experts model with 295 billion total and 21 billion active parameters and up to 256K context, released with open weights under Apache 2.0. Tencent says it is used across WorkBuddy/CodeBuddy, Yuanbao, Marvis and ima.
| Property | K-EXAONE 2.0 | Hy3 |
|---|---|---|
| Released | Jul 31, 2026 | Jul 6, 2026 |
| Developer | LG AI Research | Tencent |
| Type | llm | llm |
| License | open-weight | open-weight |
| Context window | 262,144 | — |
| Parameters | 750B total, 37B active (Mixture of Experts) | 295B total, 21B active (MoE), plus 3.8B MTP layer |
| Developer status | — | — |
| Input $/M tokens | — | $0.132 |
| Output $/M tokens | — | $0.528 |
Bold green values indicate the winner on that dimension where comparison is meaningful (lower price wins; bigger context wins; more open license wins; newer date wins). Prices are the ones each developer lists today; a model it no longer prices has none.
Programmatic access available at /api/ai/compare?a=k-exaone-2&b=tencent-hy3. CORS is open. Citation guidance: /about/editorial-process.