A SCAM provider COMMAND CODE

I bought their GOAT 10$ plan for using deepseek v4 flash . It claims it is a 10$ + 60$ credit plan, looks very good. BUT, its fee actually is 10 times higher than what you will expect. In short: It charged me 35$ for 3.6B token(my averge cache hit rate is 97%, input verses output token = 116 : 1 ), these numbers are provided by the claude-cli reported by Command Code it self. When I raise the question in their discord, I am kicked out and banned almost immediately, and the email I sent to their support is never replied . Here is my dashboard and the analysis generated by claude(base on every cli history). I reverse-engineered Command Code's rate card from my own invoices. Input is marked up 1.2x over DeepSeek. Output is marked up 4.8x. Posting the numbers in their Discord got me banned. What Command Code publishes From their model page : Input $0.14 /M Output $0.28 /M Cache read $0.003 /M Those are DeepSeek's published V4-Flash rates, passed through unchanged. From their pricing docs , verbatim: No multiplier, no agent-mode surcharge, no per-request overhead is documented anywhere I can find. Their model page also gives a worked example: Check it against their own listed rates — 135K cached, 45K miss, 12K output: 135,000 x $0.0028/M = $0.000378 45,000 x $0.14 /M = $0.006300 12,000 x $0.28 /M = $0.003360 --------- $0.010238 -> "approximately $0.01" ✓ Their published rates are meant literally. Hold onto that example — we come back to it. My workload Two independent clients, no shared code: 563 runs, 1,034,529,366 input tokens cache hit rate 97.55% (client A 97.45%, client B 97.65%) output / input 0.864% -> about 116 input : 1 output The cache hit rate is not my calculation. It is Command Code's own API response , straight off the wire: "usage":{"inputTokens":672481,"outputTokens":18222, "cacheReadTokens":634496,"cacheWriteTokens":0} 94% on that run. Across all 563 runs it averages 97.55% and stays between 96% and 98% regardless of how the session started. Deriving what I am actually charged Step 1 — uncached input rate. Find a tiny request. Nothing can be cached, so the charge is the pure uncached rate. Three separate 97-token calls, each billed $0.0000162: $0.0000162 / 97 x 1,000,000 = $0.167 /M listed $0.14 -> 1.19x Step 2 — cached input rate. Cheapest large request — big context, tiny output, so cache dominates: 128,331 input, 87 output -> $0.00048 $0.00048 / 128,331 x 1,000,000 = $0.0037 /M listed $0.003 -> 1.23x Step 3 — output rate, solved from a full month. Dashboard: 3.6B total tokens, $35.00 charged. The "all tokens" chart breaks out Input / Output / Cache Read / Cache Write, so 3.6B is input + cache-read + output. Split with my measured ratios: input 1.814 B (cached 1.770 B, miss 44.5 M) output 15.68 M Two rates are known from steps 1–2, so solve for the third: cached 1,770 M x $0.0037/M = $ 6.55 miss 44.5 M x $0.167 /M = $ 7.43 ------ $13.98 $35.00 - $13.98 = $21.02 over 15.68 M output tokens -> $1.34 /M listed $0.28 -> 4.79x Put all three back in and the month reconciles: $6.55 + $7.43 + $21.01 = $34.99 vs $35.00 charged (0.03% off) Advertised vs charged Listed Derived from invoices input, cache hit $0.003 $0.0037 1.23x input, cache miss $0.14 $0.167 1.19x output $0.28 $1.34 4.79x Now run their own worked example through both rate cards: their listed rates -> $0.0102 ("approximately $0.01") what I'm charged -> $0.0241 2.35x Why the 4.79x is the whole story My output is 0.86% of my tokens and 60% of my bill. If you spot-check a provider you check the input rate, because that's where the tokens are. Input is off by ~20%, which reads like rounding. The one line that is 4.79x is the line carrying almost all the money in a cache-heavy agent workload. Same month, priced both ways: cached miss output total at listed rates $ 4.96 $6.23 $ 4.39 $15.58 actually charged $ 6.55 $7.43 $ 21.01 $34.99 2.25x overall, and 86% of the gap is the single output line. There is no pricing-change excuse DeepSeek announced new V4 pricing on August 13, 2026 , effective 16:00 UTC August 16 . My weekly allowance was already drained before August 12 — before that announcement existed, at rates that had not changed. Whatever is happening here predates any price change and cannot be explained by one. Here is the evidence they delete my post on discord and banned me(I have to create a new account to rejoin the thread to get the evidence). preview.redd.it/dk1bt3b4i5jh1.png preview.redd.it/5spms2b4i5jh1.jpg preview.redd.it/652l93b4i5jh1.jpg