Parasail: Models Intelligence, Performance & Price
This analysis is intended to support you in choosing the best model provided by Parasail for your use-case.
Most Intelligent
DeepSeek V4.1 Flash 1017 (max)
GLM-5.3 (max)
MiniMax-M3.1-FlashGLM-5.3-Flash
Qwen3.8 27B (xhigh) (FP8)Intelligence Index
Total 30 models
Fastest
gpt-oss-120b (low)gpt-oss-120b (high)GLM-5.2 (max) (NVFP4)MiniMax-M3.1-FlashGLM-5.3 (max)Output speed
Total 30 models
Lowest Price
Gemma 3 4B (FP8)Gemma 3 27BMiniMax-M3.1-FlashGLM-5.3-FlashGemma 4 26B A4BBlended price (per 1M tokens)
Total 30 models
Parasail offers 30 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 Parasail are DeepSeek V4.1 Flash 1017 (max) (32), GLM-5.3 (max) (32), and MiniMax-M3.1-Flash (29).
- For output speed, the fastest models are gpt-oss-120b (low) (293 t/s), gpt-oss-120b (high) (287 t/s), and GLM-5.2 (max) (NVFP4) (239 t/s).
- For latency, Llama 4 Maverick (FP8) (0.77s), Qwen3 Next 80B A3B (0.81s), and Qwen3 Coder Next (FP8) (0.82s) offer the lowest time to first answer token.
- For pricing, Gemma 3 4B (FP8) ($0.05), Gemma 3 27B ($0.09), and MiniMax-M3.1-Flash ($0.10) offer the lowest blended prices per 1M tokens. Prices vary up to 2.2x across models.
- For context window size, Llama 4 Maverick (FP8) (1M), DeepSeek V4 Flash (max) (FP8) (1M), and DeepSeek V4.1 Flash (max) (1M) support the largest context windows on Parasail.
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) | 1M | Open | 32 | $0.52 | 152 | 0.68 | 17.17 | 13.20 | |||
GLM-5.3 (max) | 1M | Open | 32 | $4.92 | 245 | 0.95 | 11.14 | 8.15 | |||
MiniMax-M3.1-Flash | 1M | Proprietary | 29 | $0.44 | 309 | 0.80 | 8.90 | 6.48 | |||
Qwen3.8 27B (xhigh) (FP8) | 262k | Open | 20 | $1.78 | 84 | 1.26 | 30.92 | 23.73 | |||
MiniMax-M3 (MXFP8) | 1M | Open | 19 | $1.17 | 168 | 1.22 | 16.11 | 11.92 | |||
 Kimi K2.6 | 262k | Open | -- | -- | 83 | 1.20 | 60.59 | 53.39 | |||
Qwen3.5 397B A17B | 262k | Open | -- | -- | 91 | 1.26 | 41.85 | 35.09 | |||
Kimi K2.6 (non-reasoning) (INT4) | 262k | Open | -- | -- | 78 | 1.23 | 7.66 | -- | |||
GLM-5.3-Flash | 1M | Open | -- | -- | 122 | 0.90 | 21.44 | 16.43 | |||
Gemma 3 4B (FP8) | 131k | Open | -- | -- | 99 | 1.01 | 6.05 | -- | |||
Qwen3 Next 80B A3B | 262k | Open | -- | -- | 163 | 0.79 | 3.87 | -- | |||
Gemma 3 27B | 131k | Open | -- | -- | 41 | 2.47 | 14.76 | -- | |||
Qwen3 235B 2507 | 131k | Open | -- | -- | 37 | 1.19 | 14.61 | -- | |||
Qwen3.6 35B A3B (non-reasoning) (FP8) | 262k | Open | -- | -- | 113 | 1.02 | 5.45 | -- | |||
Gemma 4 26B A4B | 256k | Open | -- | -- | 110 | 1.07 | 23.80 | 18.19 | |||
Gemma 4 26B A4B (non-reasoning) | 262k | Open | -- | -- | 115 | 1.11 | 5.46 | -- | |||
Qwen3.6 35B A3B | 262k | Open | -- | -- | 109 | 1.08 | 55.23 | 49.55 | |||
DeepSeek V4 Flash (high) (FP8) | 1.05M | Open | -- | -- | -- | -- | -- | -- | |||
 Llama 4 Maverick (FP8) | 1.05M | Open | -- | -- | 96 | 0.89 | 6.09 | -- | |||
Llama 3.3 70B (FP8) | 131k | Open | -- | -- | 81 | 2.44 | 8.62 | -- | |||
gpt-oss-120b (low) | 131k | Open | -- | -- | 301 | 0.75 | 9.04 | 6.64 | |||
Gemma 4 31B | 262k | Open | -- | -- | 96 | 1.04 | 24.37 | 18.11 | |||
Gemma 4 31B (non-reasoning) | 262k | Open | -- | -- | 93 | 1.09 | 6.47 | -- | |||
Qwen3 VL 235B A22B (FP8) | 131k | Open | -- | -- | 45 | 1.03 | 12.14 | -- | |||
DeepSeek V4 Flash (max) (FP8) | 1.05M | Open | -- | -- | 84 | 1.51 | 74.70 | 67.20 | |||
DeepSeek V4.1 Flash (max) | 1.05M | Open | -- | -- | 163 | 0.64 | 16.00 | 12.29 | |||
gpt-oss-120b (high) | 131k | Open | -- | -- | 287 | 0.70 | 9.42 | 6.98 | |||
GLM-5.2 (max) (NVFP4) | 1M | Open | -- | -- | 270 | 0.91 | 10.17 | 7.41 | |||
Kimi K3 (max) | 1.05M | Open | -- | -- | 204 | 1.05 | 13.32 | 9.81 | |||
Qwen3 Coder Next (FP8) | 262k | Open | -- | -- | 76 | 0.79 | 7.37 | -- | |||
DeepSeek V4 Flash 0731 (max) | 1.05M | Open | -- | -- | 93 | 1.27 | 28.25 | 21.59 | |||
Key definitions
Frequently Asked Questions
Common questions about Parasail
Parasail offers 30 models that we track: DeepSeek V4.1 Flash 1017 (max), GLM-5.3 (max), MiniMax-M3.1-Flash, GLM-5.3-Flash, Qwen3.8 27B (xhigh) (FP8), MiniMax-M3 (MXFP8), Kimi K2.6, Qwen3.5 397B A17B, Kimi K2.6 (non-reasoning) (INT4), Gemma 3 4B (FP8), Qwen3 Next 80B A3B, Gemma 3 27B, Qwen3 235B 2507, Qwen3.6 35B A3B (non-reasoning) (FP8), Gemma 4 26B A4B, Gemma 4 26B A4B (non-reasoning), Qwen3.6 35B A3B, Llama 4 Maverick (FP8), Llama 3.3 70B (FP8), gpt-oss-120b (low), Gemma 4 31B, Gemma 4 31B (non-reasoning), Qwen3 VL 235B A22B (FP8), DeepSeek V4 Flash (max) (FP8), DeepSeek V4.1 Flash (max), gpt-oss-120b (high), GLM-5.2 (max) (NVFP4), Kimi K3 (max), Qwen3 Coder Next (FP8), and DeepSeek V4 Flash 0731 (max).
The most intelligent model available on Parasail is DeepSeek V4.1 Flash 1017 (max) with an Intelligence Index score of 32.
The fastest model on Parasail by output speed is gpt-oss-120b (low) at 292.6 tokens per second.
The model with the lowest time to first answer token on Parasail is Llama 4 Maverick (FP8) at 0.77s. Lower latency means faster initial response time.
The most affordable model on Parasail by blended price is Gemma 3 4B (FP8) at $0.05 per 1M tokens (7:2:1 cache hit/input/output ratio).
Prices on Parasail vary up to 51x across models, from $0.05 per 1M tokens for Gemma 3 4B (FP8) to $2.31 per 1M tokens for Kimi K3 (max).
Yes, Parasail offers an OpenAI-compatible API, making it easy to switch from OpenAI or use existing OpenAI SDK integrations.
Yes, all 30 models on Parasail support JSON mode for structured output.
28 of 30 models on Parasail support function calling (tool use).
Yes, Parasail offers 18 reasoning models: DeepSeek V4.1 Flash 1017 (max), GLM-5.3 (max), MiniMax-M3.1-Flash, GLM-5.3-Flash, Qwen3.8 27B (xhigh) (FP8), MiniMax-M3 (MXFP8), Kimi K2.6, Qwen3.5 397B A17B, Gemma 4 26B A4B, Qwen3.6 35B A3B, gpt-oss-120b (low), Gemma 4 31B, DeepSeek V4 Flash (max) (FP8), DeepSeek V4.1 Flash (max), gpt-oss-120b (high), GLM-5.2 (max) (NVFP4), Kimi K3 (max), and DeepSeek V4 Flash 0731 (max). Reasoning models use extended thinking to work through complex problems before providing an answer.
Yes, all 30 models on Parasail 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 Parasail, 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.