Skip to content

4 Uncensored LLM model added, currently available for only pay-as-you-go users.See Models

How Tokens compares

Side-by-side comparisons with other gateways and direct APIs. Each one says where the alternative is the better choice, and competitor prices are dated and sourced.

Compare models by measured speed

Measured at our gateway in Dhaka over the last 7 days, from real requests. A figure needs at least 100 requests; your own network adds to these times. Updated 23 Nov 2026, 02:39 Dhaka time.

  • Qwen 3.8 27B
    First token p50
    489 ms
    First token p90
    1.27 s
    Success rate
    87.1%
    Output speed
    362 tokens/s
  • DeepSeek V4.1 Flash 1017 Uncensored
    First token p50
    2.00 s
    First token p90
    6.05 s
    Success rate
    100%
    Output speed
    56 tokens/s
  • GLM-5.3
    First token p50
    2.05 s
    First token p90
    10.2 s
    Success rate
    92%
    Output speed
    94 tokens/s
  • GPT-5.6 Sol
    First token p50
    2.06 s
    First token p90
    6.71 s
    Success rate
    69.5%
    Output speed
    Not enough data
  • Minimax M3.1 Flash
    First token p50
    2.72 s
    First token p90
    4.70 s
    Success rate
    100%
    Output speed
    Not enough data
  • DeepSeek V4 Flash Vision (exp)
    First token p50
    2.77 s
    First token p90
    10.8 s
    Success rate
    39.7%
    Output speed
    228 tokens/s
  • Claude Haiku 5.5
    First token p50
    2.99 s
    First token p90
    8.96 s
    Success rate
    96.9%
    Output speed
    329 tokens/s
  • GLM-5.3 Flash
    First token p50
    3.40 s
    First token p90
    15.4 s
    Success rate
    98.5%
    Output speed
    115 tokens/s
  • Kimi K2.7 Code
    First token p50
    3.54 s
    First token p90
    4.73 s
    Success rate
    40.8%
    Output speed
    Not enough data
  • Ling 3.1 Flash
    First token p50
    3.75 s
    First token p90
    20.9 s
    Success rate
    85.8%
    Output speed
    Not enough data
  • GPT-6 Luna
    First token p50
    3.79 s
    First token p90
    14.9 s
    Success rate
    90.5%
    Output speed
    126 tokens/s
  • Gemini 3.7 Flash
    First token p50
    3.90 s
    First token p90
    7.88 s
    Success rate
    58.2%
    Output speed
    Not enough data
  • MiniMax M3
    First token p50
    4.11 s
    First token p90
    6.35 s
    Success rate
    98.8%
    Output speed
    117 tokens/s
  • Qwen 3.8 Flash
    First token p50
    4.15 s
    First token p90
    7.00 s
    Success rate
    98.5%
    Output speed
    87 tokens/s
  • Muse Spark 1.3 Contributor
    First token p50
    4.22 s
    First token p90
    9.60 s
    Success rate
    81.4%
    Output speed
    235 tokens/s
  • Claude Sonnet 5.5
    First token p50
    4.28 s
    First token p90
    10.4 s
    Success rate
    80.8%
    Output speed
    277 tokens/s
  • Claude Opus 5.5
    First token p50
    4.45 s
    First token p90
    13.7 s
    Success rate
    87.6%
    Output speed
    237 tokens/s
  • DeepSeek V4.1 Flash 1017
    First token p50
    4.78 s
    First token p90
    19.1 s
    Success rate
    98.3%
    Output speed
    165 tokens/s
  • Qwen 3.8 Omni Flash
    First token p50
    5.03 s
    First token p90
    18.7 s
    Success rate
    94.2%
    Output speed
    88 tokens/s
  • Gemini 3.8 Flash
    First token p50
    5.09 s
    First token p90
    10.2 s
    Success rate
    51%
    Output speed
    Not enough data
  • MiMo V2.6 Flash
    First token p50
    5.53 s
    First token p90
    14.7 s
    Success rate
    94.1%
    Output speed
    114 tokens/s
  • MiMo V2.6 Pro
    First token p50
    6.22 s
    First token p90
    14.0 s
    Success rate
    96.4%
    Output speed
    63 tokens/s
  • Hy4 Flash
    First token p50
    7.15 s
    First token p90
    10.2 s
    Success rate
    98.8%
    Output speed
    181 tokens/s
  • Qwen 3.7 Plus
    First token p50
    9.17 s
    First token p90
    23.5 s
    Success rate
    91.1%
    Output speed
    Not enough data
  • DeepSeek V4 Flash (latest)
    First token p50
    Not enough data
    First token p90
    Not enough data
    Success rate
    0%
    Output speed
    Not enough data
  • Qwen 3.7 Flash
    First token p50
    Not enough data
    First token p90
    Not enough data
    Success rate
    77%
    Output speed
    Not enough data

Comparing models instead?

Pick up to three models in the catalog and see their prices, context and measured speed side by side.

Open the model catalog