$ diff claude-opus-5-5 k2-horizon
Claude Opus 5.5 vs K2 Horizon
Values come from Claude Opus 5.5 and K2 Horizon, where each is cited to its source. This page states no benchmark result and ranks neither model — it puts two published specifications next to each other. Where a lab has not published a figure, the row says so rather than guessing.
AnthropicClaude Opus 5.5
- context
- 1M
- weights
- closed
- $/M in
- $4
- $/M out
- $20
- context
- 524K
- weights
- open
- $/M in
- —
- $/M out
- —
What actually differs
- Context window
- Claude Opus 5.5 takes 1M against 524K — modestly more room in a single request.
- Weights
- K2 Horizon publishes weights (Apache 2.0 (weights and code)); Claude Opus 5.5 is API-only. That decides self-hosting, air-gapped deployment and fine-tuning before any capability question does.
- Recency
- Claude Opus 5.5 shipped 19 days after K2 Horizon (2026-09-22 vs 2026-09-03).
Full spec
| Attribute | Claude Opus 5.5 | K2 Horizon |
|---|---|---|
| Developer | Anthropic | Ifm |
| Released | 2026-09-22 | 2026-09-03 |
| Context window | 1M tokens | 524,288 (flagship 375B-A23B, native from midtraining onward) |
| Pricing | $4/M input · $20/M output · cache reads $0.20/M · cache writes $5/M | not recorded |
| License | proprietary (API-only; no weight release) | Apache 2.0 (weights and code) |
| Availability | Claude apps, Claude Code, Claude API as `claude-opus-5-5`, Amazon Bedrock, Google Cloud Vertex AI, Microsoft Foundry | Hugging Face (weights, code, checkpoints, training data where redistributable) |