The Tool Desk
Outbyte PC Repair FREEClear out junk files and repair common Windows errorsFree Scan →Outbyte Driver Updater FREEScan for outdated or missing drivers - takes under a minuteDriver Scan →iTechGuides is reader-supported. When you buy through links on our site, we may earn an affiliate commission. As an Amazon Associate I earn from qualifying purchases. Learn more
Fujitsu and Cohere announced a strategic partnership on July 16, 2024, to jointly develop Takane, an enterprise large language model based on Cohere’s Command R+ and enhanced for Japanese-language business use. Fujitsu announced Takane’s launch on September 30, 2024, positioning it for private enterprise environments and distribution through Fujitsu Kozuchi and Data Intelligence PaaS.
What Fujitsu and Cohere announced
The July 2024 agreement combined Cohere’s language-model expertise with Fujitsu’s Japanese-language training and fine-tuning work. Fujitsu said it would be the exclusive global provider of services jointly developed through the partnership and disclosed that it had invested in Cohere. The companies also described plans for private-cloud deployment. Fujitsu and Cohere’s July 16, 2024 announcement sets out the partnership and its intended enterprise focus.
The announcement was a development and distribution agreement, not a claim that Fujitsu created a model wholly independent of Cohere. The resulting model is called Takane.
Free tools Windows power users keep installed
One-click scans. No signup required.
What Takane is and how it is intended to work
Fujitsu describes Takane as based on Cohere Command R+, with additional work for Japanese-language enterprise tasks. The companies point to Japanese training and fine-tuning, and Fujitsu’s knowledge-graph extended retrieval-augmented generation (RAG) and AI-auditing technologies. RAG retrieves relevant information from sources such as company documents to help ground model responses; it can mitigate hallucinations, but does not eliminate them.
#1 Best Overall
Fujitsu says customers can specialize the model for their operations using fine-tuning and company data. Its stated use cases include finance, government, manufacturing, research and development, healthcare, and law—areas where organizations may handle sensitive data or specialist workflows. The technology’s suitability depends on the particular data, integration, and controls involved; the announcements do not establish that Takane automatically meets every sector’s compliance requirements.
Launch date and Fujitsu delivery channels
Fujitsu announced Takane’s launch on September 30, 2024, describing it as globally available through its enterprise services. The named routes are Fujitsu Kozuchi, its AI service, and Data Intelligence PaaS, offered as part of Fujitsu Uvance. The release describes private-environment deployment as the product proposition; the security of any real deployment depends on its architecture, configuration, controls, and contract.
Fujitsu’s launch announcement provides the date and channel details: Fujitsu’s September 30, 2024 Takane launch release.
Recommended Free Tools
What Fujitsu’s benchmark figures show—and what they do not
In its September 2024 launch release, Fujitsu reported a JGLUE average of 0.92 for Takane. In the same company-published table, it listed Command R+ at 0.84, GPT-4 at 0.84, GPT-4o at 0.88, and Sonnet 3.5 at 0.86. Fujitsu and Cohere also reported Takane results on the Nejumi LLM Leaderboard 3: 0.862 for semantic understanding and 0.773 for syntactic analysis.
Fujitsu’s table gives these individual JGLUE task results:
| JGLUE task | Reported result | Measure |
|---|---|---|
| JSTS | 0.93 | Pearson |
| JCoLA | 0.84 | Balanced accuracy |
| JNLI | 0.94 | Balanced accuracy |
| JCommonsenseQA | 0.98 | Exact match |
| JSQuAD | 0.93 | Accuracy |
These are historical results measured by Fujitsu and Cohere in September 2024, not independent certification or current rankings. Fujitsu noted that multiple annotators corrected the JNLI and JCoLA ground-truth data. JGLUE is a Japanese-language version of GLUE; the release cites work by Kentaro Kurihara, Daisuke Kawahara, and Tomohide Shibata presented at the Language Processing Society of Japan annual meeting in 2022. Scores on these evaluations cannot predict performance on a particular company’s documents, languages, or workflows.
Rank #4
- Learning Japanese Workbook for Beginners: Hiragana Katakana And Kanji Quick and Easy Way to Learn the Basic Japanese Upto 300 Pages
- ABIS BOOK
- Independently published
For a buyer comparing enterprise models, benchmark scores are only one input. Compare models using the same evaluation date and Japanese tasks, and assess language coverage, deployment location and data controls, grounding on internal information, customization and integration effort, governance and audit features, and measured performance on your own workload. The company-reported table alone is not a buyer evaluation.
What the customer and company statements establish
Fujitsu’s release quotes Mizuho Financial Group Operating Officer Takefumi Yamamoto describing earlier generative-AI trials in system development and maintenance. He said, “We believe ‘Takane’ will be a valuable tool in achieving this goal,” referring to improving the quality and resilience of those processes. The release does not quantify gains achieved with Takane after launch.
Cohere’s customer story describes workflows including document processing, extraction into structured formats, reasoning, summarization, and sentiment analysis. Those are capabilities presented in the vendor’s account, not independently verified performance results. Cohere’s Fujitsu customer story provides that account.
The partnership’s central proposition is a Japanese-focused enterprise model delivered through Fujitsu’s services, with options for private deployment, customization, and grounding in company information. The announcements support that product positioning and report benchmark results; they do not establish production return on investment or guarantee security, compliance, or accuracy for every deployment.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
