Is SID-2 state of the art?Pending benchmarks.Announced Nov 7, 2026; no independent results yet.
SID AI's second-generation agentic search model. Official docs: it searches a caller's data through tools the caller implements, reads results, adapts, and returns ranked document IDs, on the first-party SID API (model sid-2) via the OpenAI Responses API. Official models page: 256k context, reasoning levels low, medium, and high, and list pricing of $0.25 input, $0.025 cached input, and $1.00 output per 1M tokens. Official technical report dated November 7, 2026. Separate from SID-2 Swarm. Not an LLM for general chat.
The details so far
- Lab
- SID AI
- Announced
- Nov 7, 2026
- API
- Live · sid-2
- Open weights
- Closed · Not released
- Task
- Agentic search: return a ranked list of relevant document IDs (official models page and docs)
- Context
- 256k tokens (official SID models page)
- Reasoning
- Low, medium, and high (official models page)
- Availability
- Public first-party SID API at https://api.sid-1.com/v1, model sid-2, with a self-serve API key (official installation and making-a-request docs, filed under research preview). Also listed on Baseten as sid/sid-2-preview and as an AWS Marketplace SageMaker package
- Price
- $0.25 input, $0.025 cached input, and $1.00 output per 1M tokens (official SID models page)
- Related
- Separate from SID-2 Swarm, which official pages describe as a separately trained multi-agent model with its own price and API
Tags marked unsure are not stated on the lab's own pages.
What would let us call it
Verdicts use only independent boards we already track. Two consecutive snapshots are required before any crown.
- Publicly announced with a source
- Usable by the public (API or weights)
- Listed on an independent board we track
- Enough results for a rank in a category
- Holds across two snapshots (crowns only)
What the lab claims
Self-reported. Shown for context; not used in any verdict.
Lab claim: on SID's self-reported 506-question benchmark across 19 corpora, SID-2 at medium effort scored 0.585 NDCG with an estimated 3.11 seconds of model inference per query (official technical report, Tables 2 and 4; SID-2 scores are the mean of three runs). The same official tables list Claude Opus 5.5 at medium effort at 0.577 NDCG and 25.1 seconds.
Sources
LABSID-2 Technical Report: Scaling Search Swarmssid.ai/research/sid-2-technical-report · Nov 7, 2026LABSID models: SID-2, SID-2 Swarm and SID-1sid.ai/modelsLABSID-2 overview (official docs)platform.sid.ai/docs/research-preview/sid-2LABSID-2 installation and API keys (official docs)platform.sid.ai/docs/research-preview/sid-2/installationLABSID-2 making a request (official docs)platform.sid.ai/docs/research-preview/sid-2/making-a-requestTracking continues until an independent board lists it.