DevPass vs OpenCode Go vs OpenCode Zen: Pricing Compared
OpenCode Go vs OpenCode Zen vs DevPass, compared on what you actually pay: monthly price, included usage per model, short-window caps, and per-token rates for the same models. Includes how each plan funds the usage it includes.
By LLM Gateway
Three ways to pay for AI coding models from the same corner of the market: OpenCode Go (a cheap subscription for open models), OpenCode Zen (pay as you go at cost), and DevPass from LLM Gateway (a subscription with one allowance across every model). The pricing pages use different units, so this post puts them on the same scale: dollars of usage per dollar paid, the caps that decide how fast you can spend it, and per-token rates for the same models, wherever each service offers them.
All figures were checked on October 10, 2026 against opencode.ai/docs/go, opencode.ai/docs/zen and the LLM Gateway model catalog. DevPass figures are the ones that take effect on October 15, 2026: new subscriptions start at 2× the plan price, existing subscriptions keep 3× until their first renewal on or after that date, and the daily and premium weekly caps apply to everyone from October 15, calculated on the new 2× allowance.
OpenCode Go vs OpenCode Zen vs DevPass at a glance
| OpenCode Go | OpenCode Zen | DevPass | |
|---|---|---|---|
| Price | $10/mo (Go), $40/mo (Go Plus) | No subscription; prepaid balance, $20 default | $29 Lite, $79 Pro, $179 Max per month |
| Included usage | A separate monthly limit per model | What you top up | 2× the plan price ($58, $158, $358) from October 15 for new subscriptions; existing ones keep 3× until their first renewal after that date. Shared by all models |
| Short-window caps | 20% per 5 hours, 50% per week, per model | Optional monthly spend limits | Daily cap 8–10%; premium models 10–15% per week |
| Fees | None on top | 4.4% + $0.30 per card transaction | None on top |
| Claude, GPT Sol, Gemini | No (Claude Haiku 5.5 and GPT Luna only) | Yes | Yes |
| Over the limit | Wait, or draw from a Zen balance | Auto-reload | Wait, or opt into pay as you go. A Reset Pass ($5 / $15 / $45; Max includes 2 per cycle) restores only the premium weekly cap |
What OpenCode Go includes per dollar
Go doesn't give one monthly pool. Each model has its own monthly limit, and the 5-hour and weekly windows are 20% and 50% of that limit. The limit depends on how cheaply OpenCode can serve the model:
| Model | Go ($10) | Go Plus ($40) |
|---|---|---|
| DeepSeek V4.1 Flash | $60 (6×) | $120 (3×) |
| GLM-5.3-Flash | $60 (6×) | $180 (4.5×) |
| MiniMax M2.7 | $60 (6×) | $240 (6×) |
| GLM-5.3 | $15 (1.5×) | $120 (3×) |
| Kimi K3 | $15 (1.5×) | $60 (1.5×) |
| Qwen3.8 Max | $15 (1.5×) | $60 (1.5×) |
On the cheapest open models, Go's $10 plan is the best ratio of any flat coding plan we track. On the expensive open models it is 1.5×, and it has no Claude Opus or Sonnet, no GPT Sol and no Gemini at any price.
DevPass works the other way: every tier includes one allowance shared by all models. New subscriptions from October 15 get 2× the plan price; existing subscriptions keep 3× until their first renewal on or after that date. Spend it all on DeepSeek, all on Claude Opus 5.5, or any mix. A daily cap (8% of the monthly allowance on Lite, 9% on Pro, 10% on Max) and a weekly cap on premium models (10%, 12%, 15%) limit how fast it goes. Both caps apply to every subscriber from October 15.
Per-token prices for the same models
When usage is metered in dollars, the per-token rate decides how many tokens a dollar buys. Rates per million tokens, input / output:
| Model | LLM Gateway (lowest listed) | OpenCode Zen | OpenCode Go |
|---|---|---|---|
| GLM-5.3-Flash | $0.07 / $0.20 | $0.15 / $0.50 | $0.15 / $0.50 |
| GLM-5.3 | $0.82 / $2.77 | $1.40 / $4.40 | $1.40 / $4.40 |
| DeepSeek V4.1 Flash | $0.12 / $0.47 | $0.30 / $1.20 | $0.15 / $0.60 off-peak, $0.30 / $1.20 peak |
| DeepSeek V4 Pro | $0.435 / $0.87 | $1.74 / $3.48 | $0.66 / $1.98 off-peak, $1.32 / $3.96 peak |
| MiniMax M2.7 | $0.08 / $0.32 | $0.30 / $1.20 | $0.30 / $1.20 |
| Gemini 3.8 Flash | $0.75 / $3.75 | $1.50 / $7.50 | Not offered |
| Kimi K3 | $2.80 / $14.13 | $3.00 / $15.00 | $3.00 / $15.00 |
| Qwen3.8 Max | $1.60 / $4.80 | $2.00 / $6.00 | $2.00 / $6.00 |
| Claude Opus 5.5 | $4.00 / $20.00 | $4.00 / $20.00 | Not offered |
| Claude Haiku 5.5 | $0.10 / $0.50 | $0.10 / $0.50 | $0.10 / $0.50 |
The gap comes from routing. Zen deliberately keeps each model on providers it has benchmarked and doesn't route to cheaper ones. LLM Gateway lists 11–13 providers for popular open models and routes each request on price, uptime and speed, so the lowest rate is usually available. Routing can pick a pricier provider when the cheapest one is slow or down, so treat our column as the floor, not a guarantee.
Carriers on LLM Gateway can also offer discounts, which is where the 20% off Qwen3.8 Max and Kimi K2.7 Code comes from. On frontier models the rates match. Zen also has about a dozen free models for a limited time (MiMo-V2.6-Flash, Muse Spark 1.3 and several stealth models), which neither Go nor DevPass prices at zero.
Every frontier model. One flat rate.
Run Claude, GPT-5.5, Gemini, and 200+ more through Claude Code, Cursor, or whatever tool you already use. From $29/mo — no token math, no surprise invoices.
How each plan pays for the usage it includes
A plan that includes more usage than its price is betting on two things: buying tokens below the public rate, and subscribers using less than their full allowance on average. All three products are open about which of those they lean on.
- Zen doesn't make the bet. OpenCode sells tokens at cost and passes card fees through, so there's no markup to fund extra usage. You pay 1× plus about 6% in fees on a $20 top-up.
- Go makes it explicitly. OpenCode's docs say the higher limits come from "bulk discounts and reserved GPU capacity", and that models without a negotiated discount get lower limits, which is why Kimi K3 and Qwen3.8 Max stop at 1.5×. Its 5-hour and weekly windows, peak pricing on DeepSeek, request for a stable session ID (used for routing and prompt caching) and one subscriber per workspace all keep heavy usage within what those discounts cover.
- DevPass draws on LLM Gateway's multi-provider routing and providers listed through Airside, which keep open-model rates at or below anyone else's. Frontier models have no such discount, which is why DevPass moves from 3× to 2× on October 15 and adds a daily cap: a shared allowance that includes Claude and GPT has to stay sustainable when people spend it on Claude and GPT.
None of this is a criticism. Caps and multipliers are how a flat price stays flat. What matters to you is which caps you'll actually hit.
Which one to pick
- Choose OpenCode Go if you mostly code with cheap open models (DeepSeek V4.1 Flash, GLM-5.3-Flash, MiniMax, MiMo) and your month fits inside its per-model limits. Nothing beats $10 for $60 on those models.
- Choose OpenCode Zen if your usage is light or irregular, or you want frontier models with no subscription. Budget for card fees on each top-up.
- Choose DevPass if you switch between open and frontier models in the same week, want one allowance instead of a limit per model, or want open models at the lowest rate across providers. At LLM Gateway's lowest GLM-5.3 rate, a dollar buys about 60–70% more tokens than at Zen or Go's rate.
For model-by-model detail, see DevPass vs OpenCode Go and DevPass vs OpenCode Zen.
- Try DevPass: $29/mo for $58 of usage across open and frontier models
- Browse model prices: live per-provider rates for every model
- Read how routing picks a provider: why open-model rates come in lower
Every frontier model. One flat rate.
Run Claude, GPT-5.5, Gemini, and 200+ more through Claude Code, Cursor, or whatever tool you already use. From $29/mo — no token math, no surprise invoices.
Frequently asked questions
- What is the difference between OpenCode Go and OpenCode Zen?
- Go is a subscription: $10/month (Go) or $40/month (Go Plus), with a separate monthly dollar limit for each model plus 5-hour (20%) and weekly (50%) windows. Zen is pay as you go: you prepay a balance, usually $20, and are charged per token at cost, plus card fees of 4.4% + $0.30 per transaction.
- Is OpenCode Go cheaper than DevPass?
- On entry price, yes: Go starts at $10 and DevPass at $29. Per dollar, Go includes up to 6× on its cheapest open models and 1.5× on premium ones like Kimi K3 or Qwen3.8 Max. From October 15, 2026, DevPass includes 2× its price as one allowance shared by every model, including Claude, GPT and Gemini frontier models that Go doesn't offer.
- Are OpenCode Zen's token prices the same as LLM Gateway's?
- For some, yes: Claude Opus, Sonnet and Haiku and MiniMax M3 are listed at the same rates on both. For open models served by several providers, LLM Gateway's lowest listed rate is lower: about 6% on Kimi K3, 20% on Qwen3.8 Max and Kimi K2.7 Code, 37–41% on GLM-5.3, 53–60% on GLM-5.3-Flash, about 60% on DeepSeek V4.1 Flash and 73% on MiniMax M2.7. Zen keeps each model on providers it has benchmarked rather than routing to the cheapest one.
- How can a plan include more usage than its price?
- Through discounts the operator gets below public rates, and because most subscribers use less than their full allowance. OpenCode says Go's higher limits come from bulk discounts and reserved GPU capacity. DevPass draws on LLM Gateway's multi-provider routing and Airside carriers. Short-window caps on both plans keep any one user from spending the whole month in a day.