SiliconFlow: Models Intelligence, Performance & Price
This analysis is intended to support you in choosing the best model provided by SiliconFlow for your use-case.
Most Intelligent
DeepSeek V4.1 Flash 1017 (max)
MiniMax-M3.1-Flash
GLM-5.3-FlashDeepSeek V4 Pro 0813 (max)
Qwen3.8 27B (xhigh)Intelligence Index
Total 39 models
Fastest
DeepSeek V4.1 Flash 1017 (max)DeepSeek V4 Flash Vision (max)DeepSeek V4.1 Flash (max)MiniMax-M3.1-FlashGLM-5.2 (max) (FP8)Output speed
Total 39 models
Lowest Price
DeepSeek V4 Flash (max) (FP8)DeepSeek V4.1 Flash 1017 (max)DeepSeek V4.1 Flash (max)MiniMax-M3.1-FlashGLM-5.3-FlashBlended price (per 1M tokens)
Total 39 models
SiliconFlow offers 39 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 SiliconFlow are DeepSeek V4.1 Flash 1017 (max) (32), MiniMax-M3.1-Flash (29), and GLM-5.3-Flash (28).
- For output speed, the fastest models are DeepSeek V4.1 Flash 1017 (max) (227 t/s), DeepSeek V4 Flash Vision (max) (171 t/s), and DeepSeek V4.1 Flash (max) (142 t/s). Speed varies significantly across models, with a 79% difference between the fastest and slowest.
- For latency, GLM-5.1 (non-reasoning) (FP8) (1.16s), GLM-5.2 (non-reasoning) (FP8) (1.21s), and Kimi K2.6 (non-reasoning) (FP8) (1.58s) offer the lowest time to first answer token.
- For pricing, DeepSeek V4 Flash (max) (FP8) ($0.07), DeepSeek V4.1 Flash 1017 (max) ($0.09), and DeepSeek V4.1 Flash (max) ($0.09) offer the lowest blended prices per 1M tokens.
- For context window size, DeepSeek V4 Pro (max) (FP8) (1M), DeepSeek V4 Flash (max) (FP8) (1M), and GLM-5.2 (non-reasoning) (FP8) (1M) support the largest context windows on SiliconFlow.
- DeepSeek V4.1 Flash 1017 (max) offers the best combination of intelligence and speed. For cost optimization, DeepSeek V4 Flash (max) (FP8) provides the most competitive pricing.
Intelligence Evaluations
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
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
Cache Behaviour
Cache Hit Rate
Cost per Task vs. Cache Hit Rate
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 | |||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
DeepSeek V4.1 Flash 1017 (max) | 1.05M | Open | 32 | $0.83 | 227 | 0.84 | 11.87 | 8.83 | |||
MiniMax-M3.1-Flash | 1M | Open | 29 | $0.68 | 130 | 3.45 | 22.74 | 15.43 | |||
GLM-5.3-Flash | 1M | Open | 28 | $1.03 | 44 | 2.62 | 58.97 | 45.08 | |||
DeepSeek V4 Pro 0813 (max) | 1.05M | Open | 23 | $2.52 | 57 | 1.80 | 45.42 | 34.90 | |||
Qwen3.8 27B (xhigh) | 262k | Open | 20 | $2.73 | 27 | 3.54 | 96.87 | 74.66 | |||
MiniMax-M3 (FP8) | 1M | Open | 19 | $1.14 | 103 | 1.52 | 25.83 | 19.45 | |||
Qwen3.8 27B (non-reasoning) | 262k | Open | 10 | $3.40 | 44 | 3.45 | 14.80 | -- | |||
 Gemma 4 12B (non-reasoning) | 262k | Open | -- | -- | 118 | 2.28 | 6.52 | -- | |||
 Kimi K2.6 (FP8) | 262k | Open | -- | -- | 27 | 1.55 | 188.10 | 167.71 | |||
Kimi K2.6 (non-reasoning) (FP8) | 262k | Open | -- | -- | 26 | 1.58 | 20.99 | -- | |||
Qwen3.5 27B (FP8) | 262k | Open | -- | -- | 24 | 3.32 | 108.03 | 83.77 | |||
Qwen3.5 122B A10B (FP8) | 262k | Open | -- | -- | 68 | 1.90 | 38.75 | 29.48 | |||
Qwen3.5 9B (FP8) | 262k | Open | -- | -- | 32 | 2.35 | 79.45 | 61.68 | |||
GLM-5 (FP8) | 200k | Open | -- | -- | -- | -- | -- | -- | |||
GLM-4.6V (non-reasoning) | 128k | Open | -- | -- | -- | -- | -- | -- | |||
Qwen3.5 35B A3B (FP8) | 262k | Open | -- | -- | 63 | 1.47 | 40.87 | 31.52 | |||
GLM-5.1 (FP8) | 205k | Open | -- | -- | 89 | 1.00 | 49.03 | 42.44 | |||
 LongCat 2.0 (FP8) | 262k | Open | -- | -- | 77 | 2.64 | 35.00 | 25.89 | |||
 Seed-OSS-36B-Instruct | 262k | Open | -- | -- | 38 | 2.93 | 68.22 | 52.23 | |||
GLM-4.5-Air | 98.3k | Open | -- | -- | 77 | 2.42 | 35.09 | 26.14 | |||
DeepSeek V4 Pro (high) (FP8) | 1.05M | Open | -- | -- | -- | -- | -- | -- | |||
DeepSeek V3.2 (non-reasoning) (FP8) | 164k | Open | -- | -- | 22 | 2.51 | 25.28 | -- | |||
Gemma 4 26B A4B (non-reasoning) (FP8) | 262k | Open | -- | -- | 90 | 2.06 | 7.58 | -- | |||
DeepSeek V4 Flash Vision (max) | 1M | Proprietary | -- | -- | 171 | 1.26 | 15.86 | 11.68 | |||
Qwen3.6 35B A3B (FP8) | 262k | Open | -- | -- | 72 | 1.21 | 82.94 | 74.80 | |||
 Ling-flash-2.0 | 131k | Open | -- | -- | 5 | 2.48 | 94.85 | -- | |||
DeepSeek V4 Flash (high) (FP8) | 1.05M | Open | -- | -- | -- | -- | -- | -- | |||
Qwen3.6 27B (FP8) | 262k | Open | -- | -- | 38 | 3.17 | 167.48 | 151.01 | |||
GLM-5.1 (non-reasoning) (FP8) | 205k | Open | -- | -- | 86 | 1.16 | 6.97 | -- | |||
Kimi K2.5 (FP8) | 262k | Open | -- | -- | 26 | 2.30 | 135.45 | 113.95 | |||
 Hy3 (FP8) | 256k | Open | -- | -- | 87 | 2.90 | 31.69 | 23.03 | |||
Qwen2.5 72B (FP8) | 32k | Open | -- | -- | 30 | 4.28 | 21.14 | -- | |||
DeepSeek V4 Pro (max) (FP8) | 1.05M | Open | -- | -- | 44 | 1.64 | 111.56 | 98.64 | |||
 Step 3.5 Flash (FP8) | 262k | Open | -- | -- | 69 | 1.59 | 38.06 | 29.18 | |||
Gemma 4 31B (FP8) | 262k | Open | -- | -- | 56 | 3.25 | 43.11 | 30.94 | |||
Gemma 4 31B (non-reasoning) (FP8) | 262k | Open | -- | -- | 61 | 3.19 | 11.45 | -- | |||
GLM-4.6V | 128k | Open | -- | -- | -- | -- | -- | -- | |||
DeepSeek V3.2 (FP8) | 164k | Open | -- | -- | 19 | 2.56 | 134.71 | 105.72 | |||
DeepSeek V4 Flash (max) (FP8) | 1.05M | Open | -- | -- | 56 | 1.47 | 111.50 | 101.04 | |||
DeepSeek V4.1 Flash (max) | 1M | Open | -- | -- | 142 | 1.93 | 19.48 | 14.04 | |||
GLM-5.2 (non-reasoning) (FP8) | 1.05M | Open | -- | -- | 109 | 1.21 | 5.80 | -- | |||
GLM-5 (non-reasoning) (FP8) | 205k | Open | -- | -- | -- | -- | -- | -- | |||
GLM-5.2 (max) (FP8) | 1.05M | Open | -- | -- | 127 | 1.19 | 20.92 | 15.78 | |||
DeepSeek V4 Flash 0731 (max) | 1.05M | Open | -- | -- | 107 | 1.93 | 25.29 | 18.69 | |||
Gemma 4 12B | 262k | Open | -- | -- | 113 | 2.22 | 24.29 | 17.65 | |||
Key definitions
Frequently Asked Questions
Common questions about SiliconFlow
SiliconFlow offers 39 models that we track: DeepSeek V4.1 Flash 1017 (max), MiniMax-M3.1-Flash, GLM-5.3-Flash, DeepSeek V4 Pro 0813 (max), Qwen3.8 27B (xhigh), MiniMax-M3 (FP8), Qwen3.8 27B (non-reasoning), Gemma 4 12B (non-reasoning), Kimi K2.6 (FP8), Kimi K2.6 (non-reasoning) (FP8), Qwen3.5 27B (FP8), Qwen3.5 122B A10B (FP8), Qwen3.5 9B (FP8), Qwen3.5 35B A3B (FP8), GLM-5.1 (FP8), LongCat 2.0 (FP8), Seed-OSS-36B-Instruct, GLM-4.5-Air, DeepSeek V3.2 (non-reasoning) (FP8), Gemma 4 26B A4B (non-reasoning) (FP8), DeepSeek V4 Flash Vision (max), Qwen3.6 35B A3B (FP8), Ling-flash-2.0, Qwen3.6 27B (FP8), GLM-5.1 (non-reasoning) (FP8), Kimi K2.5 (FP8), Hy3 (FP8), Qwen2.5 72B (FP8), DeepSeek V4 Pro (max) (FP8), Step 3.5 Flash (FP8), Gemma 4 31B (FP8), Gemma 4 31B (non-reasoning) (FP8), DeepSeek V3.2 (FP8), DeepSeek V4 Flash (max) (FP8), DeepSeek V4.1 Flash (max), GLM-5.2 (non-reasoning) (FP8), GLM-5.2 (max) (FP8), DeepSeek V4 Flash 0731 (max), and Gemma 4 12B.
The most intelligent model available on SiliconFlow is DeepSeek V4.1 Flash 1017 (max) with an Intelligence Index score of 32.
The fastest model on SiliconFlow by output speed is DeepSeek V4.1 Flash 1017 (max) at 226.6 tokens per second.
The model with the lowest time to first answer token on SiliconFlow is GLM-5.1 (non-reasoning) (FP8) at 1.16s. Lower latency means faster initial response time.
The most affordable model on SiliconFlow by blended price is DeepSeek V4 Flash (max) (FP8) at $0.07 per 1M tokens (7:2:1 cache hit/input/output ratio).
Prices on SiliconFlow vary up to 14x across models, from $0.07 per 1M tokens for DeepSeek V4 Flash (max) (FP8) to $1.03 per 1M tokens for GLM-5.1 (non-reasoning) (FP8).
Yes, SiliconFlow offers an OpenAI-compatible API, making it easy to switch from OpenAI or use existing OpenAI SDK integrations.
32 of 39 models on SiliconFlow support JSON mode for structured output.
Yes, all 39 models on SiliconFlow support function calling (tool use).
Yes, SiliconFlow offers 29 reasoning models: DeepSeek V4.1 Flash 1017 (max), MiniMax-M3.1-Flash, GLM-5.3-Flash, DeepSeek V4 Pro 0813 (max), Qwen3.8 27B (xhigh), MiniMax-M3 (FP8), Kimi K2.6 (FP8), Qwen3.5 27B (FP8), Qwen3.5 122B A10B (FP8), Qwen3.5 9B (FP8), Qwen3.5 35B A3B (FP8), GLM-5.1 (FP8), LongCat 2.0 (FP8), Seed-OSS-36B-Instruct, GLM-4.5-Air, DeepSeek V4 Flash Vision (max), Qwen3.6 35B A3B (FP8), Qwen3.6 27B (FP8), Kimi K2.5 (FP8), Hy3 (FP8), DeepSeek V4 Pro (max) (FP8), Step 3.5 Flash (FP8), Gemma 4 31B (FP8), DeepSeek V3.2 (FP8), DeepSeek V4 Flash (max) (FP8), DeepSeek V4.1 Flash (max), GLM-5.2 (max) (FP8), DeepSeek V4 Flash 0731 (max), and Gemma 4 12B. Reasoning models use extended thinking to work through complex problems before providing an answer.
Yes, 38 of 39 models on SiliconFlow are open weight models: DeepSeek V4.1 Flash 1017 (max), MiniMax-M3.1-Flash, GLM-5.3-Flash, DeepSeek V4 Pro 0813 (max), Qwen3.8 27B (xhigh), MiniMax-M3 (FP8), Qwen3.8 27B (non-reasoning), Gemma 4 12B (non-reasoning), Kimi K2.6 (FP8), Kimi K2.6 (non-reasoning) (FP8), Qwen3.5 27B (FP8), Qwen3.5 122B A10B (FP8), Qwen3.5 9B (FP8), Qwen3.5 35B A3B (FP8), GLM-5.1 (FP8), LongCat 2.0 (FP8), Seed-OSS-36B-Instruct, GLM-4.5-Air, DeepSeek V3.2 (non-reasoning) (FP8), Gemma 4 26B A4B (non-reasoning) (FP8), Qwen3.6 35B A3B (FP8), Ling-flash-2.0, Qwen3.6 27B (FP8), GLM-5.1 (non-reasoning) (FP8), Kimi K2.5 (FP8), Hy3 (FP8), Qwen2.5 72B (FP8), DeepSeek V4 Pro (max) (FP8), Step 3.5 Flash (FP8), Gemma 4 31B (FP8), Gemma 4 31B (non-reasoning) (FP8), DeepSeek V3.2 (FP8), DeepSeek V4 Flash (max) (FP8), DeepSeek V4.1 Flash (max), GLM-5.2 (non-reasoning) (FP8), GLM-5.2 (max) (FP8), DeepSeek V4 Flash 0731 (max), and Gemma 4 12B.
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 SiliconFlow, 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.