Project Tapestry - Announcements
This section contains announcements about releases and other notable updates. See also the Tapestry repository’s Releases page, announcements posted on the Project Tapestry website, and the AI Alliance blog.
Milestone Releases
Project Tapestry is managed on a three-month milestone cycle, offset by one month so we don’t have milestones ending during the holiday period at the end of December each year. Hence, this is our milestone schedule, through the end of 2027, with a summary of the goals for each past milestone and planned goals for each future milestone.
| Name | Abbrev. | Dates |
|---|---|---|
| Milestone Zero | M0 | June - August, 2026 |
| Milestone One | M1 | September - November, 2026 |
| Milestone Two | M2 | December, 2026 - February, 2027 |
| Milestone Three | M3 | March - May, 2027 |
| Milestone Four | M4 | June - August, 2027 |
| Milestone Five | M5 | September - November, 2027 |
Milestone Zero (“M0”)
| Dates | June - August, 2026 |
| Details | Milestone Zero - M0 |
| Release | v0.1.0-M0 |
| Dashboard | Tapestry Project - M0 |
Milestone “zero” (M0) was Project Tapestry’s first technical milestone. It was completed September 1, 2026. We used “zero””, rather than “one”, because M0 was about building the consortium, while also pursuing initial goals.
M0 demonstrated basic concepts of consortium training using two, geographically-distributed “sovereign” nodes (training clusters) collaborating to train a model. Other work in M0 explored the efficacy of cultural alignment techniques, started defining requirements and work groups for our ambitious data governance and management strategy, and established our development policies and practices.
More details are here.
Milestone One (“M1”)
| Dates | September - November, 2026 |
| Details | TBD |
| Release | TBD |
| Dashboard | Tapestry Project - M1 |
M1 is building on the prototype work of M0 to expand the number of sovereign nodes, increase the available compute resources, explore the feasibility and limits of heterogeneous trainings (different hardware and software stacks), beginning building cultural alignment tools and techniques, and implement the first data governance, management, and processing capabilities. A primary goal is to tune an existing open-weights model to create a very capable model targeted at a particular domain and set of use cases.
Milestone Two (“M2”)
| Dates | December, 2026 - February, 2027 |
| Details | TBD |
| Release | TBD |
| Dashboard | Tapestry Project - M1 |
In the M2 time frame, we plan to complete an end-to-end, production quality and globally-distributed training and post-training infrastructure and software stack, with further progress on building domain-specific and culturally-specific models, based on open-weight models. We also plan to complete preparation for training our own foundation models “from scratch” and to begin that process.