Frontier model comparison¶
Research cut-off date: 2026-07-22. Status: Phase 4, not yet written. This file states the scope of the work, the arguments it owns, and the research required to complete it. It contains no findings, because none have been sourced. See the changelog for phase status.
Scope¶
Compares the closed-weight models positioned by their providers at the capability frontier, on specification and on results that were produced under compatible conditions. The comparison is constrained by the verification asymmetry: for these models, parameter counts, memory footprint, and energy per token are not observable, so the comparison is necessarily thinner than the open-weight one.
Comparison rules in force¶
- Every row carries an evidence grade, a source, and a source date.
- Results produced under different harnesses, sampling policies, or tool permissions are not placed in the same ranked column without a warning adjacent to the table.
- Results whose conditions are
unstatedare never ranked against results whose conditions are known. - Provider-reported rows are labelled Grade C and are never used alone to rank one provider above another.
Research checklist¶
- Define the inclusion rule for 'frontier' explicitly and defensibly, rather than adopting provider positioning.
- Populate data/models.csv for every included model from its model card and API reference.
- Record which specification fields are 'Not publicly disclosed' rather than omitting the row.
- State in the file that Grade C rows are present and are not ranked against Grade A rows.
- Regenerate the table with
python scripts/generate_model_tables.pyand commit the result. - Run
python scripts/validate_tables.py --check-comparability.
Generated comparison¶
Generated from the data layer by scripts/generate_model_tables.py. Do not edit the region by hand.
| Provider | Model | Release date | Status | Open weights | Licence | Architecture | Total params (B) | Active params (B) | Modalities in | Context window | Max output | Reasoning mode | Tool use | Deployment | Evidence grade | Source date |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Anthropic | Claude Fable 5 | Not publicly disclosed | generally_available | no | Proprietary | Not publicly disclosed | Not publicly disclosed | Not publicly disclosed | text|image | 1000000 | 128000 | optional | function_calling | first_party_api | B | 2026-07-22 |
| Anthropic | Claude Opus 4.8 | Not publicly disclosed | generally_available | no | Proprietary | Not publicly disclosed | Not publicly disclosed | Not publicly disclosed | text|image | 1000000 | 128000 | optional | function_calling | first_party_api | B | 2026-07-22 |
| Anthropic | Claude Sonnet 5 | Not publicly disclosed | generally_available | no | Proprietary | Not publicly disclosed | Not publicly disclosed | Not publicly disclosed | text|image | 1000000 | 128000 | optional | function_calling | first_party_api | B | 2026-07-22 |
| Gemini 3.1 Flash-Lite | Not publicly disclosed | generally_available | no | Proprietary | Not publicly disclosed | Not publicly disclosed | Not publicly disclosed | text|image | Not publicly disclosed | Not publicly disclosed | optional | function_calling | first_party_api | B | 2026-07-22 | |
| Gemini 3.1 Pro | Not publicly disclosed | generally_available | no | Proprietary | Not publicly disclosed | Not publicly disclosed | Not publicly disclosed | text|image | 1000000 | 64000 | optional | function_calling | first_party_api | C | 2026-07-22 | |
| Meta | Meta Muse Spark 1.1 | Not publicly disclosed | preview | no | Proprietary | Not publicly disclosed | Not publicly disclosed | Not publicly disclosed | text|image | 1000000 | Not publicly disclosed | Not publicly disclosed | function_calling | first_party_api | B | 2026-07-22 |
| OpenAI | GPT-5.6 Luna | Not publicly disclosed | generally_available | no | Proprietary | Not publicly disclosed | Not publicly disclosed | Not publicly disclosed | text|image | 1050000 | 128000 | optional | function_calling | first_party_api | B | 2026-07-22 |
| OpenAI | GPT-5.6 Sol | Not publicly disclosed | generally_available | no | Proprietary | Not publicly disclosed | Not publicly disclosed | Not publicly disclosed | text|image | 1050000 | 128000 | optional | function_calling | first_party_api | B | 2026-07-22 |
| OpenAI | GPT-5.6 Terra | Not publicly disclosed | generally_available | no | Proprietary | Not publicly disclosed | Not publicly disclosed | Not publicly disclosed | text|image | 1050000 | 128000 | optional | function_calling | first_party_api | B | 2026-07-22 |
| xAI | Grok 4.5 | Not publicly disclosed | generally_available | no | Proprietary | Not publicly disclosed | Not publicly disclosed | Not publicly disclosed | text | 500000 | Not publicly disclosed | optional | function_calling | first_party_api | B | 2026-07-22 |