Costs

Token usage and API costs for every model run, calculated from recorded usage and model pricing.

48 total runs

$10.81Average cost per run
43m 40sAverage time per run
$3.40Median cost per run
32m 4sMedian time per run

Elapsed time and cost

Scatter chart of elapsed wall-clock time in minutes and cost in US dollars on a logarithmic scale. Lower-left means faster and cheaper. 0 30 60 90 120 150 $0.03 $0.10 $0.30 $1.00 $3.00 $10.00 $30.00 $100.00 Median 32m 4s Median $3.40 Elapsed wall-clock time (minutes) Cost (USD, logarithmic scale)

Cost by model run

Claude Fable 5 Max $81.74
Claude Fable 5.1 Max $79.04
Claude Opus 5 High $59.16
Claude Opus 4.8 Max $37.29
Devin Fusion (Fable 5.1 Max + SWE-2 High) $35.23
Claude Fable 5 High $31.23
Claude Fable 5.1 High $25.98
Claude Opus 4.8 High $18.68
Claude Fable 5.1 Low $16.46
Claude Fable 5 Low $13.78
GPT-6 Astra Max $12.93
Claude Sonnet 5 High $9.90
GPT-5.5 High $9.48
GPT-6 Astra XHigh $9.25
GPT-6 Astra High $7.22
Kimi K3 $5.77
Grok 4.5 High Fast $5.57
Kimi K2.7 Code $5.29
GPT-5.6 Sol Ultra $5.27
GPT-6 Astra Medium $5.17
GPT-5.6 Sol High $4.34
Qwen 3.7 Max $3.77
GLM 5.3 Max $3.60
GLM 5.2 Max $3.47
Grok 4.6 High $3.34
Muse Spark 1.3 xHigh $3.33
Gemini 3.8 Flash High $2.49
GPT-6 Astra Light $2.32
Composer 2.5 Fast $2.16
MiniMax M3 $1.76
Gemini 3.1 Pro $1.60
GPT-4.1 $1.45
Gemini 3.5 Flash $1.36
Muse Spark 1.3 Max Cursor $1.34
Muse Spark 1.1 High $1.13
Gemini 3.7 Flash High $1.13
GPT-5.6 Terra High $1.05
Omen Alpha High $0.87
DeepSeek V4 Pro High $0.68
Thinking Machines Inkling $0.53
GPT-5.6 Luna High $0.52
MiMo-V2.5-Pro High $0.51
GLM 5.3 Flash $0.50
DeepSeek V4.1 Flash Max $0.48
Qwen 3.7 Plus $0.45
DeepSeek V4 Flash High $0.04
MiMo-V2.5 High $0.03
SWE-2 Max $0.00

Source: local Cursor, Codex, OpenCode, and Devin CLI logs. Costs calculated with OpenUsage pricing and the rates noted below.

Fusion cost covers only the Fable 5.1 lead and the Luna compactor; the SWE-2 sidekick has no published per-token pricing and is recorded at $0.

GPT-6 Astra: $10/M input, $1/M cached input, and $50/M output.

DeepSeek bills by time of day; this run fell in the off-peak window, so off-peak rates apply (peak would double it).

SWE-2 has no published per-token pricing; run served on Devin's free launch month. Cost recorded as $0.

Elapsed time by model run

Claude Opus 5 High 2h 24m
Devin Fusion (Fable 5.1 Max + SWE-2 High) 2h 24m
Claude Fable 5 Max 2h 19m
GLM 5.3 Flash 2h 0m
Claude Fable 5.1 Max 1h 50m
GPT-5.6 Terra High 1h 33m
Claude Opus 4.8 Max 1h 25m
Kimi K3 1h 12m
Claude Opus 4.8 High 1h 12m
DeepSeek V4.1 Flash Max 1h 6m
Claude Sonnet 5 High 1h 3m
Kimi K2.7 Code 1h 2m
Claude Fable 5.1 High 1h 0m
Claude Fable 5 High 59m 25s
GLM 5.2 Max 54m 52s
Omen Alpha High 48m 20s
Grok 4.5 High Fast 43m 40s
GPT-6 Astra Max 40m 55s
Claude Fable 5.1 Low 40m 4s
GLM 5.3 Max 39m 50s
MiniMax M3 37m 33s
Grok 4.6 High 36m 28s
SWE-2 Max 35m 4s
Claude Fable 5 Low 34m 49s
GPT-6 Astra XHigh 29m 19s
Qwen 3.7 Plus 28m 42s
Muse Spark 1.3 xHigh 27m 55s
Qwen 3.7 Max 27m 6s
GPT-5.6 Sol Ultra 23m 57s
Gemini 3.7 Flash High 21m 48s
DeepSeek V4 Pro High 21m 45s
GPT-6 Astra High 21m 24s
MiMo-V2.5-Pro High 20m 10s
Gemini 3.8 Flash High 18m 39s
GPT-5.6 Sol High 15m 48s
Muse Spark 1.3 Max Cursor 14m 23s
GPT-6 Astra Medium 12m 58s
DeepSeek V4 Flash High 12m 20s
Muse Spark 1.1 High 11m 24s
Gemini 3.1 Pro 11m 13s
GPT-5.5 High 9m 55s
MiMo-V2.5 High 9m 53s
GPT-5.6 Luna High 9m 28s
Composer 2.5 Fast 8m 47s
Gemini 3.5 Flash 8m 18s
GPT-4.1 7m 56s
Thinking Machines Inkling 7m 33s
GPT-6 Astra Light 7m 8s

Wall-clock time from the first recorded run event to final recorded completion, excluding recorded pauses noted below.

Devin Fusion (Fable 5.1 Max + SWE-2 High): 12h 56m of recorded waiting time excluded.

Muse Spark 1.3 xHigh: 1h 12m of recorded waiting time excluded.

Cost concentration

$81.74Highest run
Fable 5 Max
$0.00Lowest run
SWE-2 Max
0×Highest-to-lowest spread

All runs
Claude Fable 5 Max Cursor 2h 19m 49.86M 97.9% 426k $81.74
Claude Fable 5.1 Max Cursor 1h 50m 67.75M 96.0% 608k $79.04
Claude Opus 5 High Cursor 2h 24m 75.80M 98.1% 557k $59.16
Claude Opus 4.8 Max Cursor 1h 25m 46.28M 97.9% 375k $37.29
Devin Fusion (Fable 5.1 Max + SWE-2 High) Devin 2h 24m 47.83M 93.5% 537k $35.23
Claude Fable 5 High Cursor 59m 25s 18.07M 98.0% 186k $31.23
Claude Fable 5.1 High Cursor 1h 0m 16.39M 94.3% 216k $25.98
Claude Opus 4.8 High Cursor 1h 12m 22.26M 97.6% 187k $18.68
Claude Fable 5.1 Low Cursor 40m 4s 13.04M 96.2% 148k $16.46
Claude Fable 5 Low Cursor 34m 49s 6.32M 94.7% 74k $13.78
GPT-6 Astra Max Codex 40m 55s 5.31M 96.1% 118k $12.93
Claude Sonnet 5 High Cursor 1h 3m 31.86M 97.8% 202k $9.90
GPT-5.5 High Cursor 9m 55s 3.09M 92.9% 43k $9.48
GPT-6 Astra XHigh Codex 29m 19s 3.87M 95.8% 81k $9.25
GPT-6 Astra High Codex 21m 24s 3.00M 95.2% 60k $7.22
Kimi K3 OpenCode 1h 12m 10.42M 97.2% 126k $5.77
Grok 4.5 High Fast Cursor 43m 40s 3.56M 93.5% 79k $5.57
Kimi K2.7 Code Cursor 1h 2m 21.56M 97.3% 195k $5.29
GPT-5.6 Sol Ultra Codex 23m 57s 3.64M 95.3% 91k $5.27
GPT-6 Astra Medium Codex 12m 58s 2.28M 94.4% 36k $5.17
GPT-5.6 Sol High Codex 15m 48s 3.42M 94.8% 62k $4.34
Qwen 3.7 Max OpenCode 27m 6s 5.87M 97.9% 72k $3.77
GLM 5.3 Max OpenCode 39m 50s 10.36M 97.3% 143k $3.60
GLM 5.2 Max OpenCode 54m 52s 9.56M 97.1% 149k $3.47
Grok 4.6 High Cursor 36m 28s 4.96M 95.1% 90k $3.34
Muse Spark 1.3 xHigh OpenCode 27m 55s 10.86M 89.3% 105k $3.33
Gemini 3.8 Flash High Cursor 18m 39s 8.81M 75.9% 113k $2.49
GPT-6 Astra Light Codex 7m 8s 753k 90.7% 19k $2.32
Composer 2.5 Fast Cursor 8m 47s 2.64M 97.6% 47k $2.16
MiniMax M3 OpenCode 37m 33s 25.63M 98.7% 122k $1.76
Gemini 3.1 Pro Cursor 11m 13s 2.36M 82.6% 34k $1.60
GPT-4.1 Devin 7m 56s 2.17M 94.8% 27k $1.45
Gemini 3.5 Flash Cursor 8m 18s 2.41M 86.9% 66k $1.36
Muse Spark 1.3 Max Cursor Cursor 14m 23s 5.60M 97.7% 87k $1.34
Muse Spark 1.1 High OpenCode 11m 24s 5.75M 100.0% 66k $1.13
Gemini 3.7 Flash High Cursor 21m 48s 6.29M 92.2% 90k $1.13
GPT-5.6 Terra High Codex 1h 33m 2.03M 94.8% 21k $1.05
Omen Alpha High OpenCode 48m 20s 15.57M 95.7% 221k $0.87
DeepSeek V4 Pro High OpenCode 21m 45s 8.32M 98.7% 107k $0.68
Thinking Machines Inkling OpenCode 7m 33s 1.83M 93.5% 32k $0.53
GPT-5.6 Luna High Codex 9m 28s 1.43M 87.8% 37k $0.52
MiMo-V2.5-Pro High OpenCode 20m 10s 5.71M 97.7% 60k $0.51
GLM 5.3 Flash OpenCode 2h 0m 9.65M 86.4% 121k $0.50
DeepSeek V4.1 Flash Max Devin 1h 6m 49.69M 98.4% 364k $0.48
Qwen 3.7 Plus OpenCode 28m 42s 7.25M 98.2% 63k $0.45
DeepSeek V4 Flash High OpenCode 12m 20s 4.90M 98.5% 72k $0.04
MiMo-V2.5 High OpenCode 9m 53s 2.78M 97.9% 41k $0.03
SWE-2 Max Devin 35m 4s 7.71M 95.7% 113k $0.00

Elapsed time is wall-clock time from the first recorded run event to final completion. Costs are calculated from local logs and public API model pricing. Cached input is cache-read tokens as a share of input tokens. Reasoning is included in output where reported separately.