Model Comparison

Jamba 1.5 Large vs Phi-3.5-MoE-instructWhich is better in 2026?

Jamba 1.5 Large shows notably better performance in the majority of benchmarks.

Verdict: Jamba 1.5 Large vs Phi-3.5-MoE-instruct — which is better?

Jamba 1.5 Large (by AI21 Labs) and Phi-3.5-MoE-instruct (by Microsoft) are two of the AI models people compare most. Here is how they stack up on benchmarks, price and capabilities, and which one to pick in 2026.

Jamba 1.5 Large outperforms in 5 benchmarks (ARC-C, Arena Hard, GPQA, MMLU, MMLU-Pro), while Phi-3.5-MoE-instruct is better at 2 benchmarks (GSM8k, TruthfulQA). Jamba 1.5 Large shows notably better performance in the majority of benchmarks.

Choose Jamba 1.5 Large if…

  • you want the strongest raw capability — it leads on 5 of 7 shared benchmarks

Choose Phi-3.5-MoE-instruct if…

  • you want the most recent training data — it shipped Aug 2024
Want to compare interactively?Try the playground

Performance Benchmarks

Comparative analysis across standard metrics

7 benchmarks

Jamba 1.5 Large outperforms in 5 benchmarks (ARC-C, Arena Hard, GPQA, MMLU, MMLU-Pro), while Phi-3.5-MoE-instruct is better at 2 benchmarks (GSM8k, TruthfulQA).

Jamba 1.5 Large shows notably better performance in the majority of benchmarks.

Wed Jun 24 2026 • llm-stats.com

Arena Performance

Human preference votes

Model Size

Parameter count comparison

338.0B diff

Jamba 1.5 Large has 338.0B more parameters than Phi-3.5-MoE-instruct, making it 563.3% larger.

AI21 Labs
Jamba 1.5 Large
398.0Bparameters
Microsoft
Phi-3.5-MoE-instruct
60.0Bparameters
398.0B
Jamba 1.5 Large
60.0B
Phi-3.5-MoE-instruct

Context Window

Maximum input and output token capacity

Only Jamba 1.5 Large specifies input context (256,000 tokens). Only Jamba 1.5 Large specifies output context (256,000 tokens).

AI21 Labs
Jamba 1.5 Large
Input256,000 tokens
Output256,000 tokens
Microsoft
Phi-3.5-MoE-instruct
Input- tokens
Output- tokens
Wed Jun 24 2026 • llm-stats.com

License

Usage and distribution terms

Jamba 1.5 Large is licensed under Jamba Open Model License, while Phi-3.5-MoE-instruct uses MIT.

License differences may affect how you can use these models in commercial or open-source projects.

Jamba 1.5 Large

Jamba Open Model License

Open weights

Phi-3.5-MoE-instruct

MIT

Open weights

Release Timeline

When each model was launched

Jamba 1.5 Large was released on 2024-08-22, while Phi-3.5-MoE-instruct was released on 2024-08-23.

Phi-3.5-MoE-instruct is 0 month newer than Jamba 1.5 Large.

Jamba 1.5 Large

Aug 22, 2024

1.8 years ago

Phi-3.5-MoE-instruct

Aug 23, 2024

1.8 years ago

1d newer

Knowledge Cutoff

When training data ends

Jamba 1.5 Large has a documented knowledge cutoff of 2024-03-05, while Phi-3.5-MoE-instruct's cutoff date is not specified.

We can confirm Jamba 1.5 Large's training data extends to 2024-03-05, but cannot make a direct comparison without Phi-3.5-MoE-instruct's cutoff date.

Jamba 1.5 Large

Mar 2024

Phi-3.5-MoE-instruct

—

Outputs Comparison

Notice missing or incorrect data?Start an Issue discussion→

Key Takeaways

AI21 Labs
Larger context window (256,000 tokens)
Higher ARC-C score (93.0% vs 91.0%)
Higher Arena Hard score (65.4% vs 37.9%)
Higher GPQA score (36.9% vs 36.8%)
Higher MMLU score (81.2% vs 78.9%)
Higher MMLU-Pro score (53.5% vs 45.3%)
Microsoft
Higher GSM8k score (88.7% vs 87.0%)
Higher TruthfulQA score (77.5% vs 58.3%)

Detailed Comparison

AI Model Comparison Table
Feature
AI21 Labs
Jamba 1.5 Large
Microsoft
Phi-3.5-MoE-instruct

FAQ

Common questions about Jamba 1.5 Large vs Phi-3.5-MoE-instruct.

Which is better, Jamba 1.5 Large or Phi-3.5-MoE-instruct?

Jamba 1.5 Large shows notably better performance in the majority of benchmarks. Jamba 1.5 Large is made by AI21 Labs and Phi-3.5-MoE-instruct is made by Microsoft. The best choice depends on your use case — compare their benchmark scores, pricing, and capabilities above.

How does Jamba 1.5 Large compare to Phi-3.5-MoE-instruct in benchmarks?

Jamba 1.5 Large scores ARC-C: 93.0%, GSM8k: 87.0%, MMLU: 81.2%, Arena Hard: 65.4%, TruthfulQA: 58.3%. Phi-3.5-MoE-instruct scores ARC-C: 91.0%, OpenBookQA: 89.6%, GSM8k: 88.7%, PIQA: 88.6%, RULER: 87.1%.

What are the context window sizes for Jamba 1.5 Large and Phi-3.5-MoE-instruct?

Jamba 1.5 Large supports 256K tokens and Phi-3.5-MoE-instruct supports an unknown number of tokens. A larger context window lets you process longer documents, conversations, or codebases in a single request.

What are the main differences between Jamba 1.5 Large and Phi-3.5-MoE-instruct?

Key differences include licensing (Jamba Open Model License vs MIT). See the full comparison above for benchmark-by-benchmark results.

Who makes Jamba 1.5 Large and Phi-3.5-MoE-instruct?

Jamba 1.5 Large is developed by AI21 Labs and Phi-3.5-MoE-instruct is developed by Microsoft.

Related comparisons