Compare
Claude Opus 5.5 vs Step 5 Preview: price, context and ranking
As of 29 October 2026, Claude Opus 5.5 costs $4.00 per million input tokens and $20.00 per million output tokens; Step 5 Preview costs $1.00 and $2.70. On the LMArena text leaderboard Claude Opus 5.5 scores 1,516.2 against 1,442.0 for Step 5 Preview, a gap of 74.2 points.
- Score gap
- 74.2
- Blended price
- $8.00 / $1.425
LMArena
per 1M tokens
Side by side
| Field | Claude Opus 5.5 | Step 5 Preview |
|---|---|---|
| Company | Anthropic | Stepfun |
| Input price per 1M tokens | $4.00 | $1.00 |
| Output price per 1M tokens | $20.00 | $2.70 |
| Blended price per 1M tokens (3:1) | $8.00 | $1.425 |
| Context window | 1,000,000 | 1,048,576 |
| LMArena score | 1,516.2 | 1,442.0 |
| LMArena rank | 2 | 74 |
| Arena score per dollar | 189.5 | 1,011.9 |
| Usage share, 7 days | 4.2% | 2.9% |
| Providers | 11 | 3 |
| Added to OpenRouter | 2026-09-22 | 2026-10-08 |
Frequently asked questions
Which is cheaper, Claude Opus 5.5 or Step 5 Preview?
Step 5 Preview is cheaper at a blended $1.425 per million tokens (three parts input to one part output), against $8.00 for Claude Opus 5.5.
Which scores higher on LMArena, Claude Opus 5.5 or Step 5 Preview?
Claude Opus 5.5 ranks higher on LMArena (1,516.2 against 1,442.0).
Which has the longer context window, Claude Opus 5.5 or Step 5 Preview?
Step 5 Preview has the longer context window, 1.05M tokens against 1M for Claude Opus 5.5.
Methodology and sources
The catalogue, prices, context lengths and provider counts come from OpenRouter's public Models API. They are fetched every day at 06:00 Istanbul time and stored as a daily snapshot. Weekly changes are calculated by Oesnada by comparing those snapshots; there is no data before 16 October 2026.
Usage shares are calculated from the daily token totals of OpenRouter's Data API: a model's tokens over the past 7 days divided by all tokens in the same period, including the aggregated long tail outside the top 50. Token counts come from each provider's own tokenizer and are not directly comparable across providers. The data is published by OpenRouter under CC BY 4.0. Variants such as ":free" are counted separately, so shares here can differ from OpenRouter's own ranking page.
Quality rankings come from LMArena's text (overall) leaderboard. A model is matched only when its name matches exactly; if a model has several listings, the highest-scoring one is used. "Frontier" means a model in the LMArena top 15.
"Best value" divides the LMArena score by the blended price (three parts input to one part output). Arena scores are not a linear scale, so the ratio is a rough guide. This page is produced automatically.
- OpenRouter Models API
- OpenRouter Rankings (Data API) (CC BY 4.0)
- LMArena leaderboard dataset (CC BY 4.0)
Related Oesnada stories
- Google Unveils Gemini 4 Carbon as Its Next Frontier AI Model
The model is aimed at the full defensive cycle: finding software flaws, checking whether they can be exploited, and patching them.
Importance 90/100 - GitHub Copilot Fusion, which handles model selection and orchestration, is now
The results matter because they put the cost of completed tasks, rather than the price of any single model, at the center of GitHub's pitch to software teams.
Importance 90/100 - GitHub Adds Four Million Cores of Compute to Prevent Service Outages
The surge caught the company unprepared.
Importance 90/100 - Kimi K3.1 Pro Debut Under Scrutiny Over Undisclosed Influencer Promotion Claims
The launch matters because of its mix of performance and price.
Importance 90/100
Download the data
Cite this page
Oesnada, "Claude Opus 5.5 vs Step 5 Preview: price, context and ranking", https://oesnada.com/ai-models/compare/anthropic-claude-opus-5-5-vs-stepfun-step-5-preview, accessed 29 October 2026.