[FEATURE]:Go Pro tier ($20) and Share modifier with first-month discounts

Feature hasn't been suggested before.

I have verified this feature I'm about to request hasn't been suggested before.

Describe the enhancement you want to request

Problem

I sometimes hit Go's monthly cap and the only fallback is Zen
pay-as-you-go, which is hard to budget. Separately, for personal
projects I'd be willing to share training data in exchange for
a lower price.

Proposal

Add a Pro tier (higher quota) and an "Open" variant for each
paid plan (data-sharing discount). Variant is chosen at purchase
time, switched via plan change.

	Standard	Open (data opt-in)
Go	$5 → $10/mo	$3 → $6/mo
Pro	$10 → $20/mo	$7 → $14/mo

All plans: 50% off first month
Pro quota: ~~2x Go (~~$120/mo equivalent, $24 per 5h window)
$20 ceiling on Pro keeps it within the Cursor Pro / Claude Pro range

Why "Open" as a variant, not a separate plan

Naming it Go Open / Pro Open keeps the lineup compact
(Free / Go / Pro / Zen, each with optional Open variant) and
matches OpenCode's "Open" branding. Users pick at signup;
switching uses the standard plan-change flow so pricing stays
deterministic.

Safeguards for Open variants

Auto-redaction of secrets and API keys before transmission
Limited to providers with explicit training-data agreements
Persistent UI labeling on every Open-variant session
No per-project opt-out — use Standard for sensitive work

This issue might be a duplicate of existing issues. Please check:

[FEATURE]:Opencode Go new tier #18543: Also requests new/higher OpenCode Go tiers at similar price points ($15–$20/mo)

This is what I've analyzed in just 4 hours of using OpenCode Go:

OpenRouter is 12% cheaper than the subscription's internal pricing.

Projected to ~16h/month: OpenRouter ~$53.47 vs Subscription $10 (flat rate, but throttled).

🔴 Issues

6x ROI but suffocating quotas: $61 worth for $10, but only ~16h/month.
Restrictive 5h limit: max $12.02 per 5h window. You exhausted 91% in ~3h.
Cache hits drain quota: 88% of your tokens were cache reads. Real APIs charge ~10x less for these.
GLM-5.1 at market price: $3.76 on the subscription ≈ $3.76 on OpenRouter (at par).
Misleading webpage numbers: "880-31,650 requests/model every 5h" are weighted credits, not real requests. Quota is shared across all models.

💡 Recommendation

If you use MiMo/GLM heavily: the subscription gives 6x ROI, but limited to ~16h/month. OpenRouter gives more flexibility with no quotas for ~$53.47/month.

If you only use DeepSeek: with 75% OFF, direct API is unbeatable (~$3.71/month) for this 400 requests.

After the 5 hours quota reset:

Confirmation after the reset

The continuous (5h) quota does reset.
The weekly and monthly quotas do not reset — they are cumulative and only decrease when their own cycle ends.

Actual service capacity

Concept	Value
Maximum per 5h window	$12.02 in API value
Total monthly pool	~$61
5h windows supported per month	~5 (the monthly pool runs out, not the 5h window)
Intensive usage hours per month	~15h (5 windows × 3h each)
Days of 3h sessions	~5 days, then 25 days without service

Comparison at 15h/month

(what you can actually use before the monthly quota breaks)

Service	$/month (15h)
DeepSeek promo — only DS models (what you actually used)	$1.30
DeepSeek promo — hypothetical if all traffic were DS	$3.51
$10 subscription (flat rate, throttled)	$10.00
DeepSeek without discount — only DS models	$15.91
DeepSeek without discount — all traffic	$40.69
OpenRouter (same catalog, no quotas)	$50.66

Refined conclusion

You will be able to use the service for ~15h per month (5 intensive 3h sessions). For daily developer use, you'll be without service 25 days per month.
If you switch entirely to DeepSeek with the discount, you pay $1.30/month and get unlimited access (no quotas). But you lose MiMo and GLM.
If you need MiMo/GLM, the subscription gives you a better price than OpenRouter ($10 vs $50), but with the shackle of the 15h limit.

@soyelmismo

@soyelmismo Thanks for the detailed breakdown — this is exactly the kind of
data the proposal was missing.

You're right that the real bottleneck isn't price, it's the quota structure.
A $20 Pro tier alone doesn't fix anything if the 5h window and monthly pool
scale linearly with the same throttling behavior.

Revised thinking based on your numbers:

Pro tier should rebalance the 5h window, not just double the monthly pool.
If Go gives ~3h before the 5h cap hits, Pro needs a wider window (e.g. $30+
per 5h) so intensive sessions actually complete.
Cache reads shouldn't drain quota at full rate. Your 88% cache-read
figure is the clearest waste. Billing cache hits at ~10% of normal (matching
upstream API economics) would extend usable hours dramatically without
changing the headline price.
Transparent quota display: show remaining $-value, not weighted credits.
The "880–31,650 requests" framing is misleading and your post proves it.

The "Open" (data-sharing) variant is orthogonal to this and still stands on
its own — but I agree the Pro tier proposal should be reframed around
quota mechanics, not just a higher price ceiling.

monthly pool scale linearly with the same throttling behavior.

yeah they should just put like a funding credits monthly quota, since the 5h quota isn't enough and they have weekly and monthly quota, i'm sure they can just put monthly quota and let the users burn the quota they need in one day.

* **Pro tier should rebalance the 5h window**, not just double the monthly pool.
  If Go gives ~3h before the 5h cap hits, Pro needs a wider window (e.g. $30+
  per 5h) so intensive sessions actually complete.

mm right. i agree

* **Cache reads shouldn't drain quota at full rate.** Your 88% cache-read
  figure is the clearest waste. Billing cache hits at ~10% of normal (matching
  upstream API economics) would extend usable hours dramatically without
  changing the headline price.

i can't be 100% sure that deepseek v4 pro analyzed it correctly, i have a middle proxy counting the tokens and the cached ones looks correctly measured

* **Transparent quota display**: show remaining $-value, not weighted credits.
  The "880–31,650 requests" framing is misleading and your post proves it.

the "requests" as a marketing trap is really a trap, i did like 400 requests with a lot of models and it already burned the promised 880 glm-5.1 requests in 5 hour quota. (glm is the expensive one so the other models basically don't count)

it could be better if they directly says "your 10 dollars buys you 60 dollars of funding to use all our models" but they are selling request quota in the frontpage and then counting your quota as a normal api price per token

The "Open" (data-sharing) variant is orthogonal to this and still stands on its own — but I agree the Pro tier proposal should be reframed around quota mechanics, not just a higher price ceiling.

i've already used more real tokens than i could be able to pay directly to deepseek, so i'm happy with the service. the only bad part is the way the team is selling us the service...

yep this is totally unrealistic... was it 700 "hi" requests without the harness, agent loop and system message?

@soyelmismo

@soyelmismo Yeah, I think this is the key point.

The problem is not only “we need a $20 Pro plan”. A higher-priced tier only helps if the quota mechanics scale in a way that matches real usage.

Based on your measurements, I’d reframe the proposal around three separate asks:

Expose quota as dollar-equivalent budget, not request counts

The current “requests per 5h” numbers are too easy to misunderstand because real agent usage includes system prompts, harness overhead, tool loops, cached context, and different model prices.

Showing something like:
- remaining 5h budget
- remaining weekly budget
- remaining monthly budget
- estimated burn rate
would be much clearer than marketing the plan as “X requests”.
Make the 5h window large enough for real work sessions

If Go effectively supports around a few intensive sessions per month before the monthly pool is gone, then Pro should not just be “2x monthly quota”.

A Pro tier needs a meaningfully larger 5h window, otherwise it still fails the main use case: finishing a focused coding session without hitting the short-window cap.
Charge cached reads differently from fresh input/output

I agree we can’t be 100% sure without the server-side accounting details, but if cached reads are consuming quota close to full rate, that does not match how upstream API pricing usually works.

Even a rough discount for cached tokens would make the quota feel much more aligned with actual provider costs.

So I’d summarize the revised proposal as:

add a Pro tier only if it includes a wider 5h budget, not just a larger monthly pool
display quota in dollar-equivalent terms instead of request counts
account for cached tokens separately or at a discounted rate
keep the Open/data-sharing variant as a separate pricing option, because that is orthogonal to the quota mechanics

The service may still be good value compared with paying directly through OpenRouter, but the current request-based marketing sets the wrong expectation. If quota is actually based on token-priced model cost, then the product page should say that directly.

Pessoal, o que vocês tem feito quando a cota diária/semanal/mensal acaba?
Já estou chegando no limite e não sei o que fazer. Aceito sugestões.

Atualmente utilizo o oh-my-opencode-slim com modelos diferentes para cada agente e isto tem equilibrado.
Se o plano do deepseek-v4-pro for tão bão quanto o GLM-5.1 ou KIMI-K2.7-code, eu faço a assinatura do deepseek.

+1. I'm already an OpenCode Go subscriber and constantly hit the monthly cap. I'd happily pay $50–100/month for a higher tier with proportionally larger limits. I prefer a predictable subscription over pay-as-you-go billing.

+1. Já sou assinante do OpenCode Go e constantemente atinjo o limite mensal. Pagaria com prazer de US$ 50 a US$ 100 por mês por um plano superior com limites proporcionalmente maiores. Prefiro uma assinatura previsível a um modelo de pagamento por uso.

I agree with you; I feel the same way.
I don't want to leave OpenCode, and I don't want to sign up for plans directly from OpenAI or another platform.

+1, I love my current opencode go plan but I need more, even with the new deepseek flash is not enough

I would love to switch off of my commandcode.ai $200 plan to opencode go if they had more than the go plan.

if they had more than the go plan.

i guess they are planning it. Dax is considering more plan names :p

We recently added a $40 plan

Navigation Menu

[FEATURE]:Go Pro tier ($20) and Share modifier with first-month discounts #24879

Feature hasn't been suggested before.

Describe the enhancement you want to request

Problem

Proposal

Why "Open" as a variant, not a separate plan

Safeguards for Open variants

Confirmation after the reset

Actual service capacity

Comparison at 15h/month

Refined conclusion

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

[FEATURE]:Go Pro tier ($20) and Share modifier with first-month discounts #24879

Description

Feature hasn't been suggested before.

Describe the enhancement you want to request

Problem

Proposal

Why "Open" as a variant, not a separate plan

Safeguards for Open variants

Activity

github-actions commented on Apr 29, 2026

soyelmismo commented on May 9, 2026

soyelmismo commented on May 9, 2026

Confirmation after the reset

Actual service capacity

Comparison at 15h/month

Refined conclusion

maebahesioru commented on May 14, 2026

soyelmismo commented on May 15, 2026

soyelmismo commented on May 15, 2026

maebahesioru commented on May 15, 2026

ctecservice commented on Jun 16, 2026

ChugunovRoman commented on Jun 30, 2026

ctecservice commented on Jun 30, 2026

richardmon commented on Aug 15, 2026

TheRealAlexV commented on Sep 19, 2026

soyelmismo commented on Sep 20, 2026

neriousy commented on Oct 1, 2026

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions