Partial operating KPIs
This page has 3 company-specific operating KPI charts. A sourced floor is 4. Sector membership does not count.
AI21 Funding Rounds and Valuation
| Round (date, lead) | Amount Raised | Valuation | Notable Investors |
|---|---|---|---|
| Series C first close (2023-08-31) | $155M (total raised $283M) | $1.4B | Walden Catalyst, Pitango, SCB10X, b2venture, Samsung Next, Prof. Amnon Shashua; Google and NVIDIA participating |
| Series C completion (2023-11-21) | $208M round total, adds $53M (total raised $336M) | $1.4B | New: Intel Capital, Comcast Ventures; existing: Ahren Innovation Capital |
AI21 Model-Family Release Timeline (Jurassic to Jamba)
| Date | Model / Product | Architecture | Significance |
|---|---|---|---|
| Aug 2021 | Jurassic-1 (Jumbo 178B) + AI21 Studio | Autoregressive Transformer | AI21 Studio opens in beta with Jurassic-1; the Jumbo version, at 178B parameters, is billed by AI21 as the largest language model released for general use by developers |
| 2020 | Wordtune | App layer (consumer) | AI-writing assistant; AI21 added Wordtune Read in Nov 2021 |
| Mar 2023 | Jurassic-2 (J2) + task-specific APIs | Autoregressive Transformer | Next-generation foundation models with zero-shot instruction following, lower latency and multi-language support, released with Task-Specific APIs |
| Mar 2024 | Jamba | Mamba-Transformer hybrid (SSM-Transformer) | AI21 calls it the world's first production-grade Mamba based model; 256K context, open weights under Apache 2.0 |
| Aug 2024 | Jamba 1.5 Large + Mini | Mamba-Transformer hybrid | 256K context, which AI21 says is the longest among open models and the only one backed on the RULER benchmark; on Vertex AI, Azure and NVIDIA NIM at launch, Bedrock coming soon |
| Mar 2025 | Jamba 1.6 (Large + Mini) | Mamba-Transformer hybrid | Quality and speed refresh; AI21 says Jamba Large 1.6 outperforms Mistral Large 2, Llama 3.3 70B and Command R+ on quality (Arena Hard) and speed |
| Jul 2025 | Jamba 1.7 + AI21 Maestro | Mamba-Transformer hybrid + orchestration | Jamba 1.7 (3 Jul) via AI21 Studio and Hugging Face; Maestro agentic system made available to enterprise customers (8 Jul) |
| Oct 2025 | Jamba Reasoning 3B | Mamba-Transformer hybrid (3B, on-device) | Open reasoning model under Apache 2.0 with a 256K context window that runs on laptops and phones; AI21 claims 2-5X efficiency gains over competitors such as DeepSeek, Google, Llama and Microsoft |
AI21 Strategic Backers and Model Distribution Channels
| Partner | Relationship | Distribution / Channel role | Since |
|---|---|---|---|
| NVIDIA | Series C participant | Jamba via NVIDIA NIM microservices; member of NVIDIA Inception | Series C 2023; Jamba on NIM Mar 2024 |
| Google / Alphabet | Series C participant | Jamba 1.5 family on Google Cloud Vertex AI | Series C 2023; Vertex AI Aug 2024 |
| Amazon Web Services | Distribution | Jurassic models on Amazon Bedrock; Jamba-Instruct added Jun 2024 | Apr 2023 |
| Walden Catalyst | Series C investor | No channel role stated | Series C 2023 |
| SCB10X | Series C investor | No channel role stated | Series C 2023 |
| Comcast Ventures | Series C investor (joined at completion) | No channel role stated | Nov 2023 |
| Hugging Face | Open-weights distribution | Jamba 1.5, Jamba 1.7 and Jamba Reasoning 3B open weights hosted | 2024 onward |

