01 / MODEL RESULT
GPT-6 LUNA XHigh
MODELGPT-6 LUNA · XHigh
TIME WORKED5 min 53 sec
TOKENS USEDPending
5H LIMIT USED0%
WEEKLY LIMIT USED0%
ESTIMATED COSTPending
Claude Opus 5.5 vs
GPT-6 Astra:
3D Horse Test
Thirty model results take the same low-poly 3D horse challenge in Three.js. Compare their animated results, rotate each horse, and review the available usage data.
THE CHALLENGE / 01THREE.JS · PROCEDURAL 3D · LOOPING ANIMATION
“Create a low-poly 3D horse running in a seamless loop using Three.js. Build it procedurally from primitive geometry, animate the legs, body, head, and tail, and return a single self-contained HTML file.”
RESULTS / 30
Compare thirty AI results.
One prompt. Thirty procedural horses. Rotate each result.
PROVIDER
REASONING EFFORT
02 / MODEL RESULT
GPT-6 SOL Light
MODELGPT-6 SOL · Light
TIME WORKED3 min 6 sec
TOKENS USEDPending
5H LIMIT USEDPending
WEEKLY LIMIT USEDPending
ESTIMATED COSTPending
03 / MODEL RESULT
GPT-6.1 SOL Light
MODELGPT-6.1 SOL · Light
TIME WORKED2 min 56 sec
5H LIMIT USED1%
WEEKLY LIMIT USED0%
REASONINGLight
TRIANGLES4,418
04 / MODEL RESULT
GPT-6.1 SOL Medium
MODELGPT-6.1 SOL · Medium
TIME WORKED7 min
5H LIMIT USED2%
WEEKLY LIMIT USED0%
REASONINGMedium
TRIANGLES8,928
05 / MODEL RESULT
GPT-6.1 SOL High
MODELGPT-6.1 SOL · High
TIME WORKED5 min 56 sec
5H LIMIT USED2%
WEEKLY LIMIT USED0%
REASONINGHigh
TRIANGLES3,190
06 / MODEL RESULT
GPT-6.1 SOL Very High
MODELGPT-6.1 SOL · Very High
TIME WORKED17 min 53 sec
5H LIMIT USED6%
WEEKLY LIMIT USED1%
REASONINGVery High
TRIANGLES5,352
07 / MODEL RESULT
GPT-6 SOL Very High
MODELGPT-6 SOL · Very High
TIME WORKED7 min 42 sec
5H LIMIT USED4%
The GPT-6 SOL Very High run took 7 minutes 42 seconds. The reported usage was 4% of the 5-hour limit and 0% of the weekly limit.
WEEKLY LIMIT USED0%
The GPT-6 SOL Very High run took 7 minutes 42 seconds. The reported usage was 4% of the 5-hour limit and 0% of the weekly limit.
REASONINGVery High
08 / MODEL RESULT
GPT-6 ASTRA Light
MODELGPT-6 ASTRA · Light
TIME WORKED3 min 31 sec
TOKENS USEDPending
5H LIMIT USED13%
Astra Light was measured on the ChatGPT Plus plan ($20/month): 13 percentage points of the 5-hour limit and 2 points of the weekly limit.
WEEKLY LIMIT USED2%
ESTIMATED COSTPending
09 / MODEL RESULT
GPT-6 ASTRA Medium
MODELGPT-6 Astra · Medium
TIME WORKED3 min 24 sec
TOKENS USEDUnavailable
The attached screenshot shows remaining rate limits, not token totals. Exact token usage cannot be read from it.
5H LIMIT USED52%
Before the 3 min 24 sec run, 43% of the 5-hour limit and 22% of the weekly limit were used. Afterwards, 52% and 24% were used. That is an observed increase of 9 and 2 percentage points. The screenshots do not show a token count, so exact tokens used are unavailable.
WEEKLY LIMIT USED24%
Before the 3 min 24 sec run, 43% of the 5-hour limit and 22% of the weekly limit were used. Afterwards, 52% and 24% were used. That is an observed increase of 9 and 2 percentage points. The screenshots do not show a token count, so exact tokens used are unavailable.
REASONINGMedium
10 / MODEL RESULT
GPT-6 ASTRA High
MODELGPT-6 Astra · High
TIME WORKED5 min 31 sec
TOKENS USEDUnavailable
The attached screenshot shows remaining rate limits, not token totals. Exact token usage cannot be read from it.
5H LIMIT USED43%
Before the 5 min 31 sec run, 30% of the 5-hour limit and 20% of the weekly limit were used. Afterwards, 43% and 22% were used. That is an observed increase of 13 and 2 percentage points. The screenshot does not show a token count, so exact tokens used are unavailable.
WEEKLY LIMIT USED22%
Before the 5 min 31 sec run, 30% of the 5-hour limit and 20% of the weekly limit were used. Afterwards, 43% and 22% were used. That is an observed increase of 13 and 2 percentage points. The screenshot does not show a token count, so exact tokens used are unavailable.
REASONINGHigh
11 / MODEL RESULT
GPT-6 ASTRA Very High
MODELGPT-6 Astra · Very High
TIME WORKED10 min 34 sec
TOKENS USEDUnavailable
The attached screenshot shows remaining rate limits, not token totals. Exact token usage cannot be read from it.
5H LIMIT USED150%
GPT-6 Astra Very High is listed at 150% of the 5-hour limit used and 20% of the weekly limit used. The rate-limit percentage can exceed 100%; the screenshots do not show a token count, so exact tokens used are unavailable.
WEEKLY LIMIT USED20%
GPT-6 Astra Very High is listed at 150% of the 5-hour limit used and 20% of the weekly limit used. The rate-limit percentage can exceed 100%; the screenshots do not show a token count, so exact tokens used are unavailable.
REASONINGVery High
12 / MODEL RESULT
GPT-6 ASTRA Ultra
MODELGPT-6 Astra · Ultra
TIME WORKED6 min
TOKENS USEDUnavailable
The attached screenshot shows remaining rate limits, not token totals. Exact token usage cannot be read from it.
5H LIMIT USEDNot recorded
WEEKLY LIMIT USED17%
The supplied GPT-6 Astra Ultra result took 6 minutes and used 17% of the weekly limit. No 5-hour usage percentage or token count was provided.
REASONINGUltra
13 / MODEL RESULT
GPT-6 ASTRA Pro
MODELGPT-6 Astra · Pro
CHAT COSTFree in chat with Pro subscription
REASONINGPro
14 / MODEL RESULT
GPT-5.6 LUNA Very High
MODELGPT-5.6 LUNA · Very High
TIME WORKED3 min 14 sec
TOKENS USEDUnavailable
No token count was provided for this GPT-5.6 LUNA Very High run.
5H LIMIT USED0%
The 3 min 14 sec run showed 0% used for both the 5-hour and weekly limits. Exact token usage was not provided.
WEEKLY LIMIT USED0%
The 3 min 14 sec run showed 0% used for both the 5-hour and weekly limits. Exact token usage was not provided.
REASONINGVery High
15 / MODEL RESULT
ChatGPT 5.6 SOL Instant
MODELChatGPT 5.6 SOL · Instant
TIME WORKED1 min
CHAT COSTFree in chat
NOTEFirst version was blank; revision requested
16 / MODEL RESULT
ChatGPT 5.6 SOL Medium
MODELChatGPT 5.6 SOL · Medium
CHAT COSTFree in chat
17 / MODEL RESULT
ChatGPT 5.6 SOL High
MODELChatGPT 5.6 SOL · High
TIME WORKED1 min
CHAT COSTFree in chat
18 / MODEL RESULT
Opus 5.5 Low
MODELOpus 5.5 · Low
TIME WORKED43 sec
TOKENS USED77,701
CACHE READ57,038
CACHE WRITE16,087
INPUT8
OUTPUT4,568
CURSOR PLAN USED1% · $0.20
Opus 5.5 Low used 77,701 tokens: 57,038 cache reads, 16,087 cache writes, 8 input, and 4,568 output. At Cursor’s published rates per million tokens ($0.20/$5/$4/$20 respectively), estimated token cost is $0.1832346 (about $0.18). The 1% Cursor usage equals $0.20 when compared with the $20 monthly plan price; this is a plan-usage comparison, not a separate invoice charge.
EST. TOKEN COST$0.18
19 / MODEL RESULT
Opus 5.5 Medium
MODELOpus 5.5 · Medium
TIME WORKED3 min 33 sec
TOKENS USED555,816
CACHE READ503,702
CACHE WRITE36,640
INPUT32
OUTPUT15,442
CURSOR PLAN USED3% · $0.60
Opus 5.5 totals 555,816 tokens: 503,702 cache reads, 36,640 cache writes, 32 input, and 15,442 output. At Cursor’s published rates per million tokens ($0.20/$5/$4/$20 respectively), the token estimate is $0.5929084 (about $0.59). The 3% Cursor usage equals $0.60 when compared with the $20 monthly plan price; Cursor uses a model-usage pool, so this is not a separate invoice charge.
EST. TOKEN COST$0.59
20 / MODEL RESULT
Opus 5.5 High
MODELClaude Opus 5.5 · High
TIME WORKED10 min
SUBSCRIPTIONClaude Pro · $20/month
TOKENS USEDUnavailable
The attached Claude usage screenshots show session and weekly usage percentages plus cloud credits, not token totals. Exact token usage is unavailable.
CURRENT SESSION USED56%
The Claude Pro usage screenshots show 42% of the current-session limit and 3% of the weekly limit used in one snapshot, then 56% and 5% in the other. This is an observed increase of 14 and 2 percentage points during the approximately 10-minute run. These meters report usage percentages, not token counts, so exact tokens used are unavailable.
WEEKLY LIMIT USED5%
The Claude Pro usage screenshots show 42% of the current-session limit and 3% of the weekly limit used in one snapshot, then 56% and 5% in the other. This is an observed increase of 14 and 2 percentage points during the approximately 10-minute run. These meters report usage percentages, not token counts, so exact tokens used are unavailable.
REASONINGHigh
21 / MODEL RESULT
Opus 5.5 Extra
MODELOpus 5.5 · Extra
TIME WORKED35 min
Run log label: 22:11. The time references were given in reverse chronological order. Interpreting them as started 54 minutes before the note and finished 19 minutes before it gives an elapsed time of 35 minutes.
SUBSCRIPTIONClaude Pro · $20/month
TOKENS USEDUnavailable
The attached Claude usage screenshots show session and weekly usage percentages plus cloud credits, not token totals. Exact token usage is unavailable.
CURRENT SESSION USED89%
The before-run Claude Pro screenshot showed 56% of the current-session limit and 5% of the weekly limit used. The after-run values supplied were 89% and 7%, increases of 33 and 2 percentage points. These meters are not token counts, so exact tokens used are unavailable.
WEEKLY LIMIT USED7%
The before-run Claude Pro screenshot showed 56% of the current-session limit and 5% of the weekly limit used. The after-run values supplied were 89% and 7%, increases of 33 and 2 percentage points. These meters are not token counts, so exact tokens used are unavailable.
REASONINGExtra
22 / MODEL RESULT
Opus 5.5 Max
MODELOpus 5.5 · Max
TIME WORKED1 hour
TOKENS USED28,296,992
CACHE READ26,822,994
CACHE WRITE1,127,433
INPUT162
OUTPUT346,403
CURSOR PLAN USED80% · $16.00
Opus 5.5 Max used 28,296,992 tokens across four runs: 26,822,994 cache reads, 1,127,433 cache writes, 162 input, and 346,403 output. The third run’s breakdown sums to 25,677,521 tokens. At Cursor’s standard Opus 5.5 rates per million tokens ($0.20/$5/$4/$20 respectively), the estimate is $17.9304718 (about $17.93). The 80% usage is about $16 when compared with the $20 monthly plan price; this is a plan-usage comparison, not a separate invoice charge. If this was legacy Max Mode, Cursor’s 20% surcharge would bring the estimate to about $21.52.
EST. TOKEN COST$17.93
23 / MODEL RESULT
Sonnet 5 Extra
MODELSonnet 5 · Extra
TIME WORKED5 min
CHAT COSTFree in chat
24 / MODEL RESULT
Sonnet 5.5 Low
MODELSonnet 5.5 · Low
TIME WORKED6 min
REASONINGLow
25 / MODEL RESULT
Sonnet 5.5 Medium
MODELSonnet 5.5 · Medium
TIME WORKED10 min
REASONINGMedium
26 / MODEL RESULT
Sonnet 5.5 High
MODELSonnet 5.5 · High
TIME WORKED7 min
REASONINGHigh
27 / MODEL RESULT
Sonnet 5.5 Extra
MODELSonnet 5.5 · Extra
TIME WORKED16 min
REASONINGExtra
28 / MODEL RESULT
Grok 4.7 High
MODELGrok 4.7 · High
TIME WORKED3 min 29 sec
TOKENS USED877,249
CACHE READ773,760
INPUT92,891
OUTPUT10,598
ESTIMATED COST$1.27
Grok 4.7 High used 877,249 tokens: 773,760 cache read, 0 cache write, 92,891 input, and 10,598 output. Prompt tokens (cache read + input) are 866,651, above the 200k threshold, so the long-context xAI API rate applies to the whole request: $1 per 1M cached input, $4 per 1M input, and $12 per 1M output. That is $0.7738 + $0.3716 + $0.1272 = $1.27.
29 / MODEL RESULT
DeepSeek 4 Deep Thinking
MODELDeepSeek 4 · Deep Thinking
TIME WORKED115 sec
CHAT COSTFree in chat
30 / MODEL RESULT
Gemini 3.8 Flash High
MODELGemini 3.8 Flash · High
TIME WORKED9 min 23 sec
TOKENS USED3,360,439
CACHE READ3,072,437
CACHE WRITE0
INPUT256,894
OUTPUT31,108
REASONINGHigh Effort
CURSOR PLAN USED2% · $0.40
Gemini 3.8 Flash High took 9 min 23 sec across two recorded runs: 3,279,531 + 80,908 = 3,360,439 tokens (3,072,437 cache reads, 0 cache writes, 256,894 input, and 31,108 output). At Cursor’s listed rates of $0.075 per million cache-read tokens, $0.75 per million input tokens, and $3.50 per million output tokens, the estimated token cost is $0.531981 (about $0.53). Two percent of the $20 monthly plan price is $0.40 as a plan-usage comparison; this is not a separate invoice charge.
EST. TOKEN COST$0.53