Together AI: Models Intelligence, Performance & Price
This analysis is intended to support you in choosing the best model provided by Together AI for your use-case.
Most Intelligent
Kimi K3.1 Pro
DeepSeek V4.1 Flash 1017 (max)
GLM-5.3 (max)
MiniMax-M3.1-FlashGLM-5.3-FlashIntelligence Index
Total 17 models
Lowest Time per Task
Beam 1.0 (NVFP4)
Nemotron 3.5 Super VL (NVFP4)
Muse Glimmer (high)DeepSeek V4.1 Flash 1017 (max)Muse Spark Open (xhigh)Time per Intelligence Index task (minutes)
Total 17 models
Lowest Price
Nemotron 3.5 Super VL (NVFP4)MiniMax-M3.1-FlashGLM-5.3-Flash
Qwen3.5 9B (non-reasoning) (FP8)Qwen3.5 9B (FP8)Blended price (per 1M tokens)
Total 17 models
Together AI offers 17 models, each with different intelligence, performance, and pricing characteristics. Below is a comparison of the key metrics across models.
- For intelligence, the top models on Together AI are Kimi K3.1 Pro (46), DeepSeek V4.1 Flash 1017 (max) (33), and GLM-5.3 (max) (32).
- For time per task, Beam 1.0 (NVFP4) (2.1 min), Nemotron 3.5 Super VL (NVFP4) (4.6 min), and Muse Glimmer (high) (4.8 min) finish an Intelligence Index task fastest on Together AI. Time per task varies significantly across models, with the slowest taking 3.8x longer than the quickest.
- For latency, Qwen3.5 9B (non-reasoning) (FP8) (1.02s), gpt-oss-120b (low) (2.44s), and Beam 1.0 (NVFP4) (6.70s) offer the lowest time to first answer token.
- For pricing, Nemotron 3.5 Super VL (NVFP4) ($0.09), MiniMax-M3.1-Flash ($0.10), and GLM-5.3-Flash ($0.10) offer the lowest blended prices per 1M tokens.
- For context window size, Kimi K3.1 Pro (1M), DeepSeek V4.1 Flash 1017 (max) (1M), and MiniMax-M3.1-Flash (1M) support the largest context windows on Together AI.
Intelligence
Artificial Analysis Intelligence Index
Evaluations
Agentic knowledge work, Elo
Knowledge
1 - hallucination rate
Agentic SaaS workflows
Agentic scientific research workflows in a terminal
Agentic software engineering in open-source repositories
Reasoning & knowledge
Professional document reasoning, All-pass
Agentic coding & terminal use
Physics reasoning
Quantitative analysis on spreadsheets & documents
Legal agentic work, Hallucination-Gated All-Pass Rate
Agentic business operations
Agentic real-world work tasks, Elo
Agentic interactive 3D experiences, Elo
Product and brand videos designed in code, Elo
Kubernetes incident root-cause analysis
Medical long context reasoning
Visual reasoning
Rich, varied, well-crafted language
Well-structured, scannable, answer-first
Truthful, acknowledges uncertainty, non-sycophantic
Use of filler and generic language
Intelligence Index vs. Price
Pricing
Cost per Intelligence Index Task
Intelligence Index vs. Price
Speed
Average of the single-request and multi-turn agentic workloads measured
Time per Intelligence Index Task
Quality
Endpoint Accuracy Index
Latency
Average of the single-request and multi-turn agentic workloads measured
Time to First Answer Token
Reliability
Share of requests the endpoint failed, across all the single-request and multi-turn agentic workloads measured
Endpoint Error Rate
Caching
Cache Hit Rate
Cost per Task vs. Cache Hit Rate
Further Analysis | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|
Kimi K3.1 Pro | 46 | $3.59 | 16.9 | 79.39 | 0.0% | 1M | Open | |||
DeepSeek V4.1 Flash 1017 (max) | 33 | $0.78 | 7.9 | 29.82 | 0.0% | 1M | Open | |||
GLM-5.3 (max) | 32 | $4.92 | 15.2 | 41.90 | 0.8% | 1M | Open | |||
MiniMax-M3.1-Flash | 29 | $0.46 | 17.4 | 78.24 | 9.9% | 1M | Open | |||
GLM-5.3-Flash | 28 | $0.69 | 12.5 | 24.88 | 0.7% | 1M | Open | |||
Muse Spark Open (xhigh) | 25 | $2.02 | 7.9 | 10.81 | 0.0% | 1M | Open | |||
Beam 1.0 (NVFP4) | 18 | $0.28 | 2.1 | 6.43 | 0.0% | 1M | Open | |||
Nemotron 3.5 Super VL (NVFP4) | 14 | $0.61 | 4.8 | 11.17 | 0.0% | 262k | Open | |||
 Inkling (xhigh) | 12 | $2.74 | 9.0 | 59.55 | 7.3% | 524k | Open | |||
Muse Glimmer (high) | 6 | $0.23 | 4.8 | 17.11 | 0.0% | 131k | Open | |||
 gpt-oss-120b (high) | 4 | $0.35 | 16.6 | 42.67 | 8.2% | 131k | Open | |||
DeepSeek R1 (Jan) | -- | -- | -- | -- | -- | 164k | Open | |||
MiniMax-M2.5 (FP4) | -- | -- | -- | -- | -- | 197k | Open | |||
Qwen3.5 9B (non-reasoning) (FP8) | -- | -- | -- | 1.06 | 2.1% | 262k | Open | |||
Qwen3.5 9B (FP8) | -- | -- | -- | 16.98 | 0.0% | 262k | Open | |||
GLM-4.5-Air (FP8) | -- | -- | -- | -- | -- | 131k | Open | |||
GLM-4.6 | -- | -- | -- | -- | -- | 203k | Open | |||
GLM-4.7 | -- | -- | -- | -- | -- | 203k | Open | |||
GLM-4.7 (non-reasoning) | -- | -- | -- | -- | -- | 203k | Open | |||
GLM-4.6 (non-reasoning) | -- | -- | -- | -- | -- | 203k | Open | |||
DeepSeek R1 0528 | -- | -- | -- | -- | -- | 164k | Open | |||
 Apriel-v1.5-15B-Thinker | -- | -- | -- | -- | -- | 131k | Open | |||
gpt-oss-120b (low) | -- | -- | -- | 2.41 | 8.2% | 131k | Open | |||
DeepSeek V4.1 Flash (max) | -- | -- | -- | 28.00 | 0.0% | 1M | Open | |||
GLM-5.2 (max) | -- | -- | -- | 8.71 | 0.0% | 262k | Open | |||
Kimi K3 (max) | -- | -- | -- | 27.50 | 0.0% | 1M | Open | |||
DeepSeek V3.1 (non-reasoning) | -- | -- | -- | -- | -- | 131k | Open | |||
Key definitions
Frequently Asked Questions
Common questions about Together AI
Together AI offers 17 models that we track: Kimi K3.1 Pro, DeepSeek V4.1 Flash 1017 (max), GLM-5.3 (max), MiniMax-M3.1-Flash, GLM-5.3-Flash, Muse Spark Open (xhigh), Beam 1.0 (NVFP4), Nemotron 3.5 Super VL (NVFP4), Inkling (xhigh), Muse Glimmer (high), gpt-oss-120b (high), Qwen3.5 9B (non-reasoning) (FP8), Qwen3.5 9B (FP8), gpt-oss-120b (low), DeepSeek V4.1 Flash (max), GLM-5.2 (max), and Kimi K3 (max).
The most intelligent model available on Together AI is Kimi K3.1 Pro with an Intelligence Index score of 46.
The model with the lowest time per task on Together AI is Beam 1.0 (NVFP4) at 2.1 minutes per Intelligence Index task. Time per task is the output tokens per task divided by end-to-end output speed, so lower means a task finishes sooner.
The model with the lowest time to first answer token on Together AI is Qwen3.5 9B (non-reasoning) (FP8) at 1.02s. Lower latency means faster initial response time.
The most affordable model on Together AI by blended price is Nemotron 3.5 Super VL (NVFP4) at $0.094 per 1M tokens (7:2:1 cache hit/input/output ratio).
Prices on Together AI vary up to 24.6x across models, from $0.094 per 1M tokens for Nemotron 3.5 Super VL (NVFP4) to $2.31 per 1M tokens for Kimi K3 (max).
Yes, Together AI offers an OpenAI-compatible API, making it easy to switch from OpenAI or use existing OpenAI SDK integrations.
Yes, all 17 models on Together AI support JSON mode for structured output.
Yes, all 17 models on Together AI support function calling (tool use).
Yes, Together AI offers 16 reasoning models: Kimi K3.1 Pro, DeepSeek V4.1 Flash 1017 (max), GLM-5.3 (max), MiniMax-M3.1-Flash, GLM-5.3-Flash, Muse Spark Open (xhigh), Beam 1.0 (NVFP4), Nemotron 3.5 Super VL (NVFP4), Inkling (xhigh), Muse Glimmer (high), gpt-oss-120b (high), Qwen3.5 9B (FP8), gpt-oss-120b (low), DeepSeek V4.1 Flash (max), GLM-5.2 (max), and Kimi K3 (max). Reasoning models use extended thinking to work through complex problems before providing an answer.
Yes, all 17 models on Together AI are open weight models.
Yes, provider performance can vary over time due to infrastructure changes, load balancing, and updates. We continuously benchmark all providers and display historical performance trends in the "Over Time" charts.
When choosing a model on Together AI, consider: intelligence (for quality-sensitive tasks), time per task (how quickly the model completes an Intelligence Index task), latency (for interactive applications requiring quick first responses), pricing (for cost-sensitive workloads), and features like context window size, JSON mode, or function calling support.