Skip to content

[FEATURE]: OpenCode Go should pass through DeepSeek V4 Pro's permanent price reduction to subscribers #29008

@linchengyuan

Description

@linchengyuan

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

Background

DeepSeek V4 Pro has undergone two price adjustments since launch:

Date Change Cache Hit (per 1M) Cache Miss (per 1M) Output (per 1M)
Apr 24 (launch) Original pricing $0.145 $1.74 $3.48
Apr 26 Cache hit reduced to 1/10 across all models $0.0145 $1.74 $3.48
Apr 26 V4 Pro 75% off promo starts $0.003625 $0.435 $0.87
May 22 75% discount made permanent $0.003625 $0.435 $0.87

The permanent price is now 1/4 of the original: $0.435/M input (cache miss) / $0.003625/M (cache hit) / $0.87/M output.

The problem

OpenCode Go's quotas are dollar-denominated ($12/5h, $30/week, $60/month). The docs state that "limits are defined in dollar value" and provide estimated request counts based on "typical Go usage patterns."

For DeepSeek V4 Pro, those estimates are:

  • 3,450 requests per 5h
  • 750 input (cache miss) + 82,000 cached + 290 output tokens per request

By reverse-calculating:

Item Per-request tokens OpenCode Go's implied price Per-request cost
Cache miss input 750 $1.74/M (original, no discount) $0.001305
Cache hit input 82,000 $0.0145/M (post-Apr-26 cache reduction, no 75% off) $0.001189
Output 290 $3.48/M (original, no discount) $0.001009
Total 83,040 ~$0.00350

3,450 requests × $0.00350 = $12.08 ≈ $12/5h ✓

Conclusion: OpenCode Go is using DeepSeek's original pre-discount pricing ($1.74 input / $3.48 output) to calculate quota consumption. Only the cache hit price was adjusted from $0.145 to $0.0145 (reflecting the Apr 26 across-the-board cache reduction). The 75% discount — now permanent since May 22 — has not been passed through.

What should be possible at the discounted rate

If Go applied DeepSeek's current permanent pricing:

  • Cost per request: ~$0.000876
  • Requests possible per 5h within $12 limit: ~13,700
  • Currently offered: 3,450
  • Users receive ~75% fewer requests than the discount would justify within the same $12 quota window

Precedent: OpenCode has adjusted before

On Apr 27, commit #24575 updated V4 Pro's estimated requests from 1,300 to 3,450 in response to DeepSeek's cache hit price change. This proves OpenCode Go can and does adjust quotas when upstream pricing changes. The current request is to do the same for the now-permanent 75% discount.

Request

I'm not asking for a free ride — $10/month for multi-model access is fair. I'm asking for transparency and alignment:

  1. Pass through the discount — update V4 Pro's quota to reflect the current permanent API pricing (~13,700/5h), consistent with how the cache hit adjustment was handled
  2. Or, explicitly state the policy — if Go is a fixed-margin subscription that does not track upstream price changes, document this clearly so users can make informed decisions
  3. At minimum, show implied per-token rates — publish the effective per-token cost Go uses for each model, not just request counts, so users understand the actual pricing

Related: https://www.80aj.com/2026/05/05/opencode-go-deepseek-discount-issue/

Metadata

Metadata

Assignees

Labels

No labels
No labels

Type

No type
No fields configured for issues without a type.

Projects

No projects

Milestone

No milestone

Relationships

None yet

Development

No branches or pull requests

Issue actions