Together AI: Models Intelligence, Performance & Price
Analysis of Together AI's models across key metrics including quality, price, output speed, latency, context window & more. This analysis is intended to support you in choosing the best model provided by Together AI for your use-case.
Most Intelligent
Intelligence index
Total 12 models
Fastest
Output speed
Total 12 models
Lowest Price
Blended price (per 1M tokens)
Total 12 models
Together AI offers 12 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 (max) (60), MiniMax-M3 (45), and Kimi K2.7 Code (43).
- For output speed, the fastest models are Kimi K2.7 Code (261 t/s), Kimi K2.6 (FP4) (146 t/s), and Nemotron 3 Ultra (142 t/s). Speed varies significantly across models, with a 171% difference between the fastest and slowest.
- For latency, Kimi K2.6 (FP4) (0.58s), Qwen3.5 9B (FP8) (0.82s), and Gemma 4 31B (FP8) (1.20s) offer the lowest time to first answer token.
- For pricing, Gemma 3n E4B ($0.07), Qwen3.5 9B (FP8) ($0.18), and Qwen3.5 9B (FP8) ($0.18) offer the lowest blended prices per 1M tokens. Prices vary up to 6.8x across models.
- For context window size, Kimi K3 (max) (1M), MiniMax-M3 (1M), and Inkling (524k) support the largest context windows on Together AI.
Highlights
Intelligence Evaluations
Artificial Analysis Intelligence Index
Intelligence Evaluations
Agentic real-world work tasks, (Elo-500)/2000
Agentic tool use
Agentic coding & terminal use
Coding
Reasoning & knowledge
Scientific reasoning
Physics reasoning
Knowledge
1 - hallucination rate
Long context reasoning
Agentic knowledge work, Elo
Agentic SaaS workflows
Legal agentic work, criterion pass rate
Agentic business operations
Instruction following
Long-horizon agentic tasks
Kubernetes incident root-cause analysis
Visual reasoning
Intelligence Index vs. Price
Context Window
Context Window
Pricing
Intelligence Index vs. Price
Performance Summary
Output Speed vs. Price
Speed
Measured by Output Speed (tokens per second)
Output Speed
Latency
Measured by Time (seconds) to First Token
Latency: Time To First Answer Token
End-to-End Response Time
Seconds to output 500 tokens, calculated based on time to first token, 'thinking' time for reasoning models, and output speed
End-to-End Response Time vs. Price
Further Analysis | |||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
Kimi K3 (max) | 1M | Open | 60 | $1.43 | 73 | 0.97 | 35.28 | 27.45 | |||
MiniMax-M3 | 1M | Open | 45 | $0.21 | 62 | 1.01 | 41.54 | 32.43 | |||
Kimi K2.7 Code | 262k | Open | 43 | $0.30 | 261 | 0.48 | 10.92 | 8.53 | |||
Inkling | 524k | Open | 42 | $0.62 | 95 | 0.60 | 27.06 | 21.16 | |||
Nemotron 3 Ultra | 512k | Open | 38 | $0.40 | 142 | 0.68 | 20.20 | 16.01 | |||
Kimi K2.6 (FP4) | 262k | Open | 35* | -- | 146 | 0.58 | 4.00 | -- | |||
MiniMax-M2.5 (FP4) | 197k | Open | 34* | -- | -- | -- | -- | -- | |||
 GLM-4.7 | 203k | Open | 34 | $0.28 | -- | -- | -- | -- | |||
Gemma 4 31B (FP8) | 262k | Open | 30 | $0.10 | 52 | 1.43 | 44.82 | 33.68 | |||
GLM-4.6 | 203k | Open | 29 | $0.33 | -- | -- | -- | -- | |||
GLM-4.7 | 203k | Open | 27* | -- | -- | -- | -- | -- | |||
GLM-4.6 | 203k | Open | 23* | -- | -- | -- | -- | -- | |||
Gemma 4 31B (FP8) | 262k | Open | 22 | $0.11 | 57 | 1.20 | 9.99 | -- | |||
Qwen3.5 9B (FP8) | 262k | Open | 22 | $0.30 | 96 | 0.78 | 26.70 | 20.74 | |||
 Apriel-v1.5-15B-Thinker | 131k | Open | 22* | -- | -- | -- | -- | -- | |||
 DeepSeek V3.1 | 131k | Open | 21* | -- | -- | -- | -- | -- | |||
Qwen3 Coder Next (FP8) | 262k | Open | 21 | $0.47 | -- | -- | -- | -- | |||
Apriel-v1.6-15B-Thinker | 131k | Open | 21* | -- | -- | -- | -- | -- | |||
Qwen3.5 9B (FP8) | 262k | Open | 21* | -- | 95 | 0.82 | 6.09 | -- | |||
DeepSeek R1 0528 | 164k | Open | 20* | -- | -- | -- | -- | -- | |||
DeepSeek R1 (Jan) | 164k | Open | 19 | $0.50 | -- | -- | -- | -- | |||
GLM-4.5-Air (FP8) | 131k | Open | 17* | -- | -- | -- | -- | -- | |||
Llama 4 Maverick | 1.05M | Open | 14 | $0.04 | -- | -- | -- | -- | |||
Llama 3.3 70B Turbo | 131k | Open | 9 | $0.16 | 102 | 1.48 | 6.37 | -- | |||
Gemma 3n E4B | 32.8k | Open | 1* | -- | 41 | 1.30 | 13.55 | -- | |||
 Cogito v2.1 | 164k | Open | -- | -- | -- | -- | -- | -- | |||
Key definitions
Frequently Asked Questions
Common questions about Together AI
Together AI offers 12 models that we track: Kimi K3 (max), MiniMax-M3, Kimi K2.7 Code, Inkling, Nemotron 3 Ultra, Kimi K2.6 (FP4), Gemma 4 31B (FP8), Gemma 4 31B (FP8), Qwen3.5 9B (FP8), Qwen3.5 9B (FP8), Llama 3.3 70B Turbo, and Gemma 3n E4B.
The most intelligent model available on Together AI is Kimi K3 (max) with an Intelligence Index score of 60.
The fastest model on Together AI by output speed is Kimi K2.7 Code at 261.4 tokens per second.
The model with the lowest time to first answer token on Together AI is Kimi K2.6 (FP4) at 0.58s. Lower latency means faster initial response time.
The most affordable model on Together AI by blended price is Gemma 3n E4B at $0.07 per 1M tokens (7:2:1 cache hit/input/output ratio).
Prices on Together AI vary up to 35x across models, from $0.07 per 1M tokens for Gemma 3n E4B 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 12 models on Together AI support JSON mode for structured output.
11 of 12 models on Together AI support function calling (tool use).
Yes, Together AI offers 7 reasoning models: Kimi K3 (max), MiniMax-M3, Kimi K2.7 Code, Inkling, Nemotron 3 Ultra, Gemma 4 31B (FP8), and Qwen3.5 9B (FP8). Reasoning models use extended thinking to work through complex problems before providing an answer.
Yes, all 12 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), output speed (for throughput-intensive tasks), 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.