Model comparison
Compare Claude Haiku 4.5 and Celeris-1 Magnus using the same provider-sourced text & reasoning rubric. No mystery score and no invented benchmark ranking.
Facts checked September 14, 2026
Set your usage. Your estimate updates as you type.
Example: 6,000 input tokens for 10 pages, plus a 500-token summary. Page lengths vary; adjust the numbers below.
One run sends your input to the model once and receives an answer. Tokens are pieces of text: input is what you send, output is the answer you receive.
Estimates exclude taxes, tools, cache storage/writes, free allowances and custom discounts. Image estimates cover output only, not prompt or reference-image charges. Quality modes differ by model. Unlisted settings are not treated as free.
| Model | Access | Estimated total (USD) |
|---|---|---|
| Claude Haiku 4.5Anthropic | Anthropic | $0.0085 |
| Celeris-1 MagnusCeleris | Celeris | No reviewed rate |
Quick take
A compact Claude model for quick replies, classification and narrowly scoped assistants.
Haiku 4.5 uses manual extended thinking, not the adaptive thinking settings of the newer Claude models.
Celeris' agent-focused diffusion language model for fast reasoning, tool loops, structured actions, and OpenAI-compatible integration.
The public price page labels Celeris-1 rather than Magnus, and the service remains early. Confirm Magnus billing, capacity, support, data retention, and a production SLA directly.
Compare the published facts
Values use each provider's own published units and limits. A blank means the provider did not publish a directly comparable value in the sources reviewed.
| Text & reasoning | Claude Haiku 4.5 | Celeris-1 Magnus |
|---|---|---|
| Context windowMaximum combined prompt and working context documented by the provider. | 200K tokens | 131,072 tokens combined |
| Maximum outputProvider-published response limit, where available. | 64K tokens | Any positive limit within the combined context |
| Knowledge cutoffLatest reliable knowledge date explicitly published by the model provider. Search and connected tools can retrieve newer information but do not change the model's built-in cutoff. | February 2025 | Not published |
| Input priceCurrent standard list price per million input tokens unless noted. | $1 / 1M | $0.20 / 1M listed family rate; verify Magnus |
| Output priceCurrent standard list price per million output tokens unless noted. | $5 / 1M | $0.70 / 1M listed family rate; verify Magnus |
| InputsMedia types accepted by the listed model endpoint. | Text, image | Text; image content parts supported by the API |
| Tools & agentsSelected native tools and agent-building capabilities, not an exhaustive list. | Tool use, agents, vision, coding workflows | Tools, JSON schema, reasoning low/medium/xhigh |
How to choose
Start with the job you need to complete, then validate cost, access, and policy details on your exact provider route.
Anthropic does not train on API data without express permission. Retention depends on the API feature, account agreement and cloud provider.
Provider and API links
Celeris says API inputs and outputs are processed to provide the service and monitor abuse and reliability. Its public notice does not give one fixed content-retention period or no-training promise; enterprise VPC options are available.
Provider and API links
Frequently asked questions
Claude Haiku 4.5: A compact Claude model for quick replies, classification and narrowly scoped assistants. Celeris-1 Magnus: Celeris' agent-focused diffusion language model for fast reasoning, tool loops, structured actions, and OpenAI-compatible integration.
Consider Claude Haiku 4.5 when your priority is Data extraction. Consider Celeris-1 Magnus when your priority is Low-latency tool-using agents. Test both with your own data and provider route before committing.
No. This comparison aligns provider-published facts for the Text & reasoning category. It does not claim a universal winner or combine incompatible third-party benchmark scores.
