GPT-5.2-Codex
Current directoryCapability, route economics, service-level measurements and lifecycle evidence in one model-specific decision record.
Published Sep 14, 2026 · stale evidence · 2 benchmarks · 2 sources
Published benchmark evidence includes benchlm content observed outside the 8-day evidence window.
- Source overall score
- 58.36
- Blended $ / 1M (10:2)
- $3.79
- TTFT p50
- Not reported
Six-axis evidence
Derived rank-percentiles · same axes and scale as Compare · throughput remains a separate runtime axis.
GPT-5.2-Codex
Exact capability values
Derived rank-percentile = 100 × (cohort size − rank) / (cohort size − 1). Only a published rank and its exact cohort size qualify; missing or single-entry cohorts remain gaps.
| Domain | Score / unit | Derived rank-percentile | Rank / cohort size | Source |
|---|---|---|---|---|
| Agentic | 53.62 score | 66.91 | 47 / 140 | BenchLM |
| Coding | 57.42 score | 70.83 | 43 / 145 | BenchLM |
| Reasoning | Not reported | Not reported | Not reported | Not reported |
| Math | Not reported | Not reported | Not reported | Not reported |
| Multimodal | Not reported | Not reported | Not reported | Not reported |
| Throughput | Not reported | Not reported | Not reported | Not reported |
Gaps are missing or ineligible evidence, never zero. Overall and Knowledge are not silently substituted for Math or runtime.
Runtime SLA evidence
A measured route and its conditions are required. Benchmark scores do not establish latency or throughput.
Time to first token (seconds)
Output throughput (tokens/second)
- TTFT
- Not reported
- Throughput
- Not reported
- Measured at
- Not reported
- Region
- Not reported
- Percentile
- Not reported
- Concurrency
- Not reported
- Streaming
- Not reported
- Runtime source
- Not reported
Identity, limits & route
- Provider
- OpenAI
- Access
- Proprietary
- Context
- 400,000
- Maximum output
- Not reported
- Maximum input
- Not reported
- Input modalities
- Not reported
- Output modalities
- Not reported
- Representative route
- benchlm:gpt-5-2-codex
- Price source
- BenchLM source · observed Aug 30, 2026
Lifecycle & sunset
- Release
- 2025-12-18
- Provider lifecycle status
- Not reported
- Sunset
- Not reported
- Replacement
- Not reported
- Directory status
- current
Directory presence is not a provider support or retirement promise.
Open the full lifecycle radarEndpoint & itemized price matrix
Source-shaped price fields stay separate from the blended scenario above.
The first matrix preserves prices recorded with these benchmark results. Current catalog facts are shown separately below and may differ.
| Host / route | Input $/1M | Output $/1M | Cache read | Cache write | Long context | Context window | Max output | Availability | Source / observed |
|---|---|---|---|---|---|---|---|---|---|
| openaibenchlm:gpt-5-2-codex | $1.75 | $14.00 | Not reported | Not reported | Not reported | 400,000 | Not reported | Route not verifiedPrice evidence: primary | BenchLM sourceAug 30, 2026 |
Current catalog · exact matching routes
Catalog catalog_e7f5de78d2108d49f3be10ea_62bbb2e3 · published Sep 14, 2026. Matches require the same offer ID, provider and source model ID. A provider route expiry is not a model retirement or proof that the endpoint is offline.
No exact current catalog route facts are reported for this profile. Historical prices above remain benchmark-pinned evidence.
No current exact binding: benchlm:gpt-5-2-codex.
Workload-aware cost example
Derived scenario · 10M input + 2M output tokens / month
- Source input price
- $1.75 / 1M
- Source output price
- $14.00 / 1M
- Benchmark-pinned monthly cost
- $45.50
- Current exact-route monthly cost
- Not reported
- Current example catalog revision
- catalog_e7f5de78d2108d49f3be10ea_62bbb2e3
- Current price observation
- Not reported
- Formula
- 10 × input price + 2 × output price
- Blended price
- (10 × input price + 2 × output price) ÷ 12
Excludes cache, retries, tool calls and long-context tiers. Unknown required prices make the example unavailable. This is a price estimate, not a verified route-availability claim.
Use your own workloadHistory, conflicts & limitations
- Reported release date
Recorded in the published model identity.
- First directory observation
Revision benchmark_3efc47868bcf9f8d4b17d35cb33ce0e7. This is not necessarily the model release date.
- Last directory observation
Revision benchmark_2f59d4658a789000c14f0d0c7a177e62.
- BenchLM evidence observed
- BenchLM evidence observed
- No route conflict is flagged in this snapshot; absence of a flag is not independent agreement.
- Cache write, long-context rates and runtime conditions remain Not reported until qualified evidence is published.
- Validate the selected route price, context limits, and evidence freshness before choosing.
Benchmark ledger and provenance
Public overall score 58.36 at source rank #87.
- Profile benchmark revision
- benchmark_2f59d4658a789000c14f0d0c7a177e62
- Publication cache revision
- benchmark_2f59d4658a789000c14f0d0c7a177e62+cache-20260914055705000-c21dcf92-e210-43a5-bb34-f9c3e2271701
- Benchmark-pinned price catalog
- catalog_e7f5de78d2108d49f3be10ea_62bbb2e3
- Current catalog context
- catalog_e7f5de78d2108d49f3be10ea_62bbb2e3
- Current catalog published
- 2026-09-14T05:27:06.588Z
- Published
- 2026-09-14T05:57:05.000Z
- Checked
- 2026-09-14T05:57:05.000Z
- Generated
- 2026-09-14T05:57:05.000Z
- Fallback
- none
- Verified alias
- Canonical profile
- Model key
- source:benchlm:gpt-5-2-codex
BenchLM · retained
- Source content observed
- 2026-08-30T02:15:58.000Z
- Source published
- 2026-08-29T16:34:49.708Z
- Last attempted
- 2026-09-14T05:58:23.098Z
- Last successfully verified
- 2026-08-30T02:15:58.000Z
- Content hash
- sha256:93e32b81e518ece89faa6ea4a32e4697d33bcb2e948c0f9d41d3bff11d7bc9eb
- Retained from
- benchmark_3b14dc55ef5307cfe91356d3ea348101
- Latest failure
- source_policy_rejected at 2026-09-14T05:58:23.098Z
LiteLLM · verified
- Source content observed
- 2026-09-14T05:58:55.316Z
- Source published
- Not reported
- Last attempted
- 2026-09-14T05:58:55.316Z
- Last successfully verified
- 2026-09-14T05:58:55.316Z
- Content hash
- sha256:b0e4eb602920d79596dc648e174fd4f58bffb9c6ca7d38269d412a66c061dd3c
- Retained from
- Not retained
- Latest failure
- No recorded failure
LMArena · verified
- Source content observed
- 2026-08-30T02:15:58.000Z
- Source published
- Not reported
- Last attempted
- 2026-09-14T06:11:06.112Z
- Last successfully verified
- 2026-09-14T06:11:06.112Z
- Content hash
- sha256:ffc0b4a062807d93eebf7040ef6fa707a807bbdbc2908a74d747c41662701de8
- Retained from
- Not retained
- Latest failure
- No recorded failure
OpenRouter · pinned
- Source content observed
- 2026-09-14T05:27:06.588Z
- Source published
- Not reported
- Last attempted
- Not reported
- Last successfully verified
- Not reported
- Content hash
- sha256:0e0ea7d964404497cd00b3a75f238e18ddd9b54b81b77ca311d4ffda54ff8e48
- Retained from
- Not retained
- Latest failure
- No recorded failure
| Benchmark | Score | Rank | Weight | Last Updated | Source |
|---|---|---|---|---|---|
| benchlm:category:agentic | 53.62 score | #47 | Not published | Aug 29, 2026 | BenchLMsupported · public-leaderboard |
| benchlm:category:coding | 57.42 score | #43 | Not published | Aug 29, 2026 | BenchLMsupported · public-leaderboard |
| benchlm:overall:raw | 58.36 score | #87 | Not published | Aug 29, 2026 | BenchLMsupported · public-leaderboard |