3 OpenCode Go
OpenCode Go is a service I recommend without reservations because it is fairly transparent about what it provides even if the naming and other aspects get confusing.
There are three services under the OpenCode umbrella:
- the coding harness: OpenCode
- the metered LLM access: OpenCode Zen
- the subscription LLM access: OpenCode Go
With Go, you pay $10/mo1 and you can use credits worth at least that much. It even supports UPI, including mandates for recurring payments.
This is where it gets interesting. Every model it offers has a different multiplier applied to the credits:
- DeepSeek V4 Flash: 6x2
- MiMo 2.5: 6x
- MuseSpark 1.2 Contributor: 6x
- GLM 5.3: 1.5x
What this means is: you can incur charges of $15-60 on each of these models during the month.3 The multiplier is used to reconcile the figure with the price you paid for the subscription:
- The 6x models: $60/6 = $10.
- The 1.5x models: $15/1.5 = $10.
This is the way to reason about it. Otherwise, you will only get confused.
Now, let us compare the two extremes. Here are the prices for two models:
| Type | MiMo 2.5 | GLM 5.3 |
|---|---|---|
| Cached Read | $0.0028/M | $0.26/M |
| Input | $0.14/M | $1.40/M |
| Output | $0.28/M | $4.40/M |
A 99% cache rate is fairly typical in coding workflows. So, the hypothetical tokens used and cost:
| Type | MiMo 2.5 | GLM 5.3 | Token Ratio |
|---|---|---|---|
| Cached Read | 10B = $28 | 50M = $13 | 200:1 |
| Input | 100M = $14 | 500K = $0.70 | 200:1 |
| Output | 50M = $14 | 250K = $1.10 | 200:1 |
| Total | $56 | $14.80 |
You are getting nearly 200 times more tokens across the board with MiMo over GLM 5.3. But you must acknowledge that MiMo is far dumber than GLM. You get what you pay for.
$5 the first month. Well, the introductory price has been withdrawn. I guess people abused it by cycling through accounts. ↩- Before the price bump. ↩
- OpenCode does have Claude-style usage limits. ↩