openai / Chat
GPT-5.2 Fast
openai/gpt-5.2-lowGPT-5.2 with low thinking — fastest replies, lowest cost
Catalog status
Enabled
Research checked
2026-10-07
Benchmark results
Results are source-attributed, not TokenBazaar measurements. Different tests and thinking variants are not interchangeable.
1 of 1 results
lineage-bench (overall, avg of lineage-8/16/32/64/128)
Reasoning · Third-party report
Test conditions and source
gpt-5.2, reasoning.effort=low, lineage-bench (graph relationship reasoning), via OpenAI API (author states repeated directly, not only via OpenRouter). Community-reproduced micro-benchmark, not an OpenAI-published table; single independent tester, small forum post, not peer reviewed. Publisher excerpt: gpt-5.2 (low) | 0.660 | 1.000 | 0.975 | 0.800 | 0.400 | 0.125
Tested model: openai/gpt-5.2-low
OpenAI Developer Community forum · independent lineage-bench results by user sszymczy · Checked 2026-10-07
Model specifications
Context window
400K
Published provider limit
Maximum output
128K
Published provider limit
Knowledge cutoff
Aug 31, 2025
Upstream model
Input and output modalities
Provider-reported formats; API compatibility is detailed below.
Input
Output
- Exact model ID
- openai/gpt-5.2-low
- Model type
- Chat
- Context window
- 400,000 tokens
- Maximum output
- 128,000 tokens
- Knowledge cutoff
- Aug 31, 2025
A context window is the total conversation budget, not a separate maximum input allowance; generated output and reasoning can use that budget. Published provider limits are not independently tested TokenBazaar request limits.
Versions and thinking levels
Choose from 4 enabled versions. The exact ID determines the version and its price; benchmark scores do not carry over between variants.
Features and tool support
TokenBazaar’s exposed interface, not every feature advertised by the provider. “Available” describes the implemented interface, not a successful test of every input or tool.
Reasoning configuration
The catalog version determines the requested reasoning effort.
Selected effort: low. Reasoning level is not a measured intelligence or speed score.
Coding and agent workflows
Text/code generation and customer-managed tool loops.
Coding benchmark results do not establish a hosted terminal, sandbox, autonomous browser or guaranteed task success.
Streaming answers
AvailableIncremental text through TokenBazaar’s chat endpoint; Claude also has native Messages access.
Customer-defined function tools
LimitedOnly customer-defined function tools are forwarded. Model/protocol compatibility applies; your application authorizes and executes every tool.
Forced or named tool selection
LimitedThe chat route accepts auto, none and required. A named-tool choice is not forwarded; exact model compatibility is not independently tested.
Images and PDFs
Not verifiedThe chat bridge accepts image/PDF content. Provider input modalities are listed separately; file size, content and model limits still apply.
Hosted web search / browsing
Not exposedNo provider-hosted web search or web-fetch tool is exposed. Customer-owned search can be implemented as an authorized function tool.
Hosted code execution / computer use
Not exposedNo built-in execution environment or computer-control tool is provided. Code generation is not code execution.
Hosted file search / persistent assistants
Not exposedNo hosted file-search index, Assistants endpoint or persistent provider agent is exposed.
Batch / fine-tuning / cached-price discounts
Not exposedNo public batch or fine-tuning endpoint, or separate cached-token discount, is offered here.
Sources and verification
Provider specifications describe the upstream model, not a guarantee of every feature through TokenBazaar. Prices come from the enabled catalog; benchmark results belong to the exact tested model.
- Provider model specifications
Vendor-reported
Last checked - OpenAI - Introducing GPT-5.2
Vendor-reported
Last checked - OpenAI Developer Community forum · independent lineage-bench results by user sszymczy
Third-party report
Last checked
Research notes (1)
- Published provider limits; maximum-capacity requests have not been independently exercised through TokenBaazar. Thinking-level variants use the same provider model; reasoning and answer tokens share the output budget.