[FEATURE]: Restore the previous 1M context window for opencode/deepseek-v4-flash-free #27929

This issue doesn't fully meet our contributing guidelines.

What needs to be fixed:

• Feature request titles must start with [FEATURE]: (e.g., [FEATURE]: Restore the previous 1M context window for opencode/deepseek-v4-flash-free)

Please edit this issue to address the above within 2 hours, or it will be automatically closed.

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

• Qwen3.6 Plus Context Limit Decreased? #22595: Similar report of a model's context window being reduced unexpectedly

If you believe this was flagged incorrectly, please let a maintainer know.

Thanks for updating the issue! It now meets the contributing guidelines and the needs:compliance label has been removed. We appreciate you taking the time to provide clear context around the impact of the context window change.

Agree, the 1m context did worked very well, if it can be brought back, would be awesome, if it is abused and uses too many resources, maybe you guys could create a new tier account, like a donator account, that if an account donated for example 50$ it can use extra content not available to free tier, one time fee or something no subscription base, at least this can generate some revenue, that otherwise will be 0 because people will stuck to free models. Daily limits are ok, but the 1m context made a difference in quality.

Many are wiling to cintribute to what we believe is a good product and fair offer.

Agree, the 1m context did worked very well, if it can be brought back, would be awesome, if it is abused and uses too many resources, maybe you guys could create a new tier account, like a donator account, that if an account donated for example 50$ it can use extra content not available to free tier, one time fee or something no subscription base, at least this can generate some revenue, that otherwise will be 0 because people will stuck to free models. Daily limits are ok, but the 1m context made a difference in quality.

Many are wiling to cintribute to what we believe is a good product and fair offer.

you can solve this for now by adding this to .config/opencode/opencode.json :

  "provider": {
    "opencode": {
      "models": {
        "deepseek-v4-flash-free": {
          "limit": {
            "context": 1048576,
            "output": 384000
          }
        }
      }
    },
  }

Agree, the 1m context did worked very well, if it can be brought back, would be awesome, if it is abused and uses too many resources, maybe you guys could create a new tier account, like a donator account, that if an account donated for example 50$ it can use extra content not available to free tier, one time fee or something no subscription base, at least this can generate some revenue, that otherwise will be 0 because people will stuck to free models. Daily limits are ok, but the 1m context made a difference in quality.
Many are wiling to cintribute to what we believe is a good product and fair offer.

you can solve this for now by adding this to .config/opencode/opencode.json :

  "provider": {
    "opencode": {
      "models": {
        "deepseek-v4-flash-free": {
          "limit": {
            "context": 1048576,
            "output": 384000
          }
        }
      }
    },
  }

But would this actually work ? since opencode deepseek-v4-flash-free context window was degraded on thier end

Agree, the 1m context did worked very well, if it can be brought back, would be awesome, if it is abused and uses too many resources, maybe you guys could create a new tier account, like a donator account, that if an account donated for example 50$ it can use extra content not available to free tier, one time fee or something no subscription base, at least this can generate some revenue, that otherwise will be 0 because people will stuck to free models. Daily limits are ok, but the 1m context made a difference in quality.
Many are wiling to cintribute to what we believe is a good product and fair offer.

you can solve this for now by adding this to .config/opencode/opencode.json :

  "provider": {
    "opencode": {
      "models": {
        "deepseek-v4-flash-free": {
          "limit": {
            "context": 1048576,
            "output": 384000
          }
        }
      }
    },
  }

But would this actually work ? since opencode deepseek-v4-flash-free context window was degraded on thier end

Yeah it works, but in my opinion it is easier to just turn off compaction and then manually compact when needed.

Agree, the 1m context did worked very well, if it can be brought back, would be awesome, if it is abused and uses too many resources, maybe you guys could create a new tier account, like a donator account, that if an account donated for example 50$ it can use extra content not available to free tier, one time fee or something no subscription base, at least this can generate some revenue, that otherwise will be 0 because people will stuck to free models. Daily limits are ok, but the 1m context made a difference in quality.
Many are wiling to cintribute to what we believe is a good product and fair offer.

you can solve this for now by adding this to .config/opencode/opencode.json :

  "provider": {
    "opencode": {
      "models": {
        "deepseek-v4-flash-free": {
          "limit": {
            "context": 1048576,
            "output": 384000
          }
        }
      }
    },
  }

But would this actually work ? since opencode deepseek-v4-flash-free context window was degraded on thier end

Yeah it works, but in my opinion it is easier to just turn off compaction and then manually compact when needed.

Yeah it worked for me, game changer ...

Agree, the 1m context did worked very well, if it can be brought back, would be awesome, if it is abused and uses too many resources, maybe you guys could create a new tier account, like a donator account, that if an account donated for example 50$ it can use extra content not available to free tier, one time fee or something no subscription base, at least this can generate some revenue, that otherwise will be 0 because people will stuck to free models. Daily limits are ok, but the 1m context made a difference in quality.
Many are wiling to cintribute to what we believe is a good product and fair offer.

you can solve this for now by adding this to .config/opencode/opencode.json :

  "provider": {
    "opencode": {
      "models": {
        "deepseek-v4-flash-free": {
          "limit": {
            "context": 1048576,
            "output": 384000
          }
        }
      }
    },
  }

Doesnt it just make you lose context? as I understand this if the context window of the model itself is 200k, increasing it in client side only will result in losing context? increasing it from client side just makes the app not comapact it at 200k or push more context but from server end it trims 800k context and recieves 200k. Please correct me if I am wrong, did any one tested it?

Agree, the 1m context did worked very well, if it can be brought back, would be awesome, if it is abused and uses too many resources, maybe you guys could create a new tier account, like a donator account, that if an account donated for example 50$ it can use extra content not available to free tier, one time fee or something no subscription base, at least this can generate some revenue, that otherwise will be 0 because people will stuck to free models. Daily limits are ok, but the 1m context made a difference in quality.
Many are wiling to cintribute to what we believe is a good product and fair offer.

you can solve this for now by adding this to .config/opencode/opencode.json :

  "provider": {
    "opencode": {
      "models": {
        "deepseek-v4-flash-free": {
          "limit": {
            "context": 1048576,
            "output": 384000
          }
        }
      }
    },
  }

Doesnt it just make you lose context? as I understand this if the context window of the model itself is 200k, increasing it in client side only will result in losing context? increasing it from client side just makes the app not comapact it at 200k or push more context but from server end it trims 800k context and recieves 200k. Please correct me if I am wrong, did any one tested it?

No try to send to the model more than 1M context it will literally tell u u cant an dit works over 200k so that means the server side accepts -1M tokens