Overview
| Name | Kosmos / Edison Analysis · aliases: Kosmos, Edison Analysis |
|---|---|
| Full name | No separate expanded name |
| Ambiguous names (not auto-linked) | Finch |
| Family | Not a family entry. Single-model entry, not a family |
| Organization | Edison Scientific, FutureHouse |
| Organization type | AI-native biotech |
| Country | United States |
| Model class | AI-scientist agent |
| Domain | Text, LLMs & agents |
| First release | † partly verified Basis: Edison Scientific announcement (with arXiv 2511.02824) Note: arXiv v1 may be 2025-11-04, earlier than this date; unverified source |
| Other milestones | |
| Latest update | · no later event Basis: Initial announcement; later items are partnership news, not new versions No later sourced event found (as of 2026-11-16) Note: 2026 partnership news with Incyte and Pfizer is not a model update source The most recent sourced event, not a claim that nothing changed since. |
| Status | Product only, no public artifact · Available only as a platform and API |
| Openness | Code: Closed Weights: Closed Not open source; available through the Edison platform and API |
| Benchmarks reported | 1 counted (self 1) |
| Wet-lab validation | None recorded |
| Unverified fields | † first_release |
| As of | 2026-11-16 |
What it does
Kosmos is positioned as an "AI scientist" that cycles between literature search and data analysis to produce cited research reports. Edison Analysis (formerly Finch) is its data-analysis agent, offered as a separate product.
Benchmarks reported
Every row is one pairing from the survey's relation file; a score belongs to the benchmark page's ranking, where it is shown with metric, eval set, source type and date. 1 of 1 pairings are counted.
| Benchmark | Relation | Eval set / protocol | Source type | Note and evidence |
|---|---|---|---|---|
| 29 BixBench | Developer report, official protocol | original BixBench | Self-reported | Self-report on the original; the product's mapping to this site entry is unclear and it is not on the successor's board †unverified · evidence |
Other evaluations (outside the 44)
None recorded.
Developers report: Source states no developer claim
Independent evidence: No independent evidence found. Only a re-run by competitor Phylo
Training data
Group in Figure 7: Public data only.
| Class | Value | Detail |
|---|---|---|
| a. Public databases | Yes | Base LLM with public literature and data; training data not disclosed |
| b. Private or licensed data | Not disclosed | |
| c. Own wet-lab data used for training | No | |
| d. Wet lab used only for validation | No |
Wet-lab validation
None. The developers report no wet-lab validation.
Background
- Organization: Edison Scientific is the for-profit spin-out of the nonprofit lab FutureHouse (founded 2023-09) and also makes LAB-Bench and BixBench
- Funding: Announced a $70M seed round on 2025-12-18, led by Triatomic Capital and Spark Capital
- Products and partnerships: Partnerships announced with Incyte (2026-05-19) and Pfizer (2026-10-29); founding subscriptions ended, new plans pending
- Related models: Robin, Biomni
Reading notes
- BixBench comes from the same company, so look for third-party re-runs
- The site BixBench cell is actually an Edison Analysis number; its identity vs Kosmos is still to be clarified
Caveats
- First-release date may be one day after arXiv v1; not re-checked
- Closed source; cannot be reproduced externally
Sources
- Official page edisonscientific.com (company blog)
- arXiv preprint 2511.02824 (preprint)
- Official page edisonscientific.com (company blog)
Corrections
None recorded.
Part of the AI for Bio benchmarks survey · all 112 models · open in the explorer. Record as of 2026-11-16.