Skip to content

feat: add qwen/qwen3.6-plus (paid) to OpenRouter and KiloCode model lists - #6599

Open
iRonin wants to merge 1 commit into
NousResearch:mainfrom
iRonin:feat/add-qwen36-plus-paid
Open

feat: add qwen/qwen3.6-plus (paid) to OpenRouter and KiloCode model lists#6599
iRonin wants to merge 1 commit into
NousResearch:mainfrom
iRonin:feat/add-qwen36-plus-paid

Conversation

@iRonin

@iRonin iRonin commented Apr 9, 2026

Copy link
Copy Markdown
Contributor

Summary

Add qwen/qwen3.6-plus (paid tier) to the model catalogs for both OpenRouter and KiloCode providers. The :free variant remains listed alongside it.

Changes

hermes_cli/models.py

  • Added qwen/qwen3.6-plus to OPENROUTER_MODELS with descriptor "1M context, thinking"
  • Added qwen/qwen3.6-plus to the kilocode provider list in _PROVIDER_MODELS

agent/model_metadata.py

  • Added "qwen3.6-plus": 1048576 to DEFAULT_CONTEXT_LENGTHS (1M context)

Model Specs

  • Context: 1,048,576 tokens (1M)
  • Thinking: Supported (extended thinking / reasoning)
  • Pricing: Fetched live from OpenRouter /v1/models endpoint
  • Availability: OpenRouter + KiloCode

Both providers support this model ID natively -- no additional normalisation needed beyond the existing qwen vendor prefix detection in model_normalize.py.

@iRonin

iRonin commented Apr 9, 2026

Copy link
Copy Markdown
Contributor Author

Closes #6601

iRonin added a commit to iRonin/hermes-agent-nous that referenced this pull request Apr 9, 2026
Preserves 58 fork feature commits aligned with v2026.4.8 upstream:
- Dual queue system (priority/standard)
- Subagent panel UI with /subagents command
- Multi-stash management
- Terminal title customization
- Per-skill model routing (model/provider/skill/skills)
- Browser tool enhancements with user-data-dir support
- Async delegation infrastructure
- Config system extensions (config.yaml, hermes_cli/config.py)

Conflict resolutions:
- delegate_tool.py: merged fork's schema (model/provider/skill/skills)
  with upstream's ACP support (acp_command/acp_args)
- cli.py: resolved Chrome candidate method naming conflict
- api_server.py: merged /v1/sessions endpoint documentation

Includes PR NousResearch#6599: Qwen 3.6-Plus support (1M context window)
iRonin added a commit to iRonin/hermes-agent-nous that referenced this pull request Apr 9, 2026
All custom features merged onto v2026.4.8 + PR NousResearch#6599 (Qwen 3.6-Plus):
- Terminal title chain (12 commits)
- Ctrl+D delete char, double ESC clear, Ctrl+P peek
- Arrow keys free cursor, Up/Down history
- Ctrl+P history pager, Ctrl+O full message toggle
- Subagent control panel (Ctrl+X)
- Multi-item stash with panel, stash_auto_restore
- Dual queue: followup (Alt+Enter) + steering dispatch
- Per-skill model routing (model/provider/skill/skills)
- Gateway: /v1/sessions, model override
- Browser: /browser connect with Chrome profile
- Session: /resume picker, session title
- Async delegation (delegate_task_async tools)

Conflicts resolved using previously-resolved versions.
@iRonin
iRonin force-pushed the feat/add-qwen36-plus-paid branch from cf040be to cc277e9 Compare April 11, 2026 18:49
Add the paid tier of qwen3.6-plus alongside the existing :free variant
to the OpenRouter model list and KiloCode provider list. Also add
1M context (1,048,576) to the default context length lookup.
@iRonin
iRonin force-pushed the feat/add-qwen36-plus-paid branch from cc277e9 to 82bb1ef Compare April 11, 2026 20:37
@alt-glitch alt-glitch added type/feature New feature or request P3 Low — cosmetic, nice to have comp/cli CLI entry point, hermes_cli/, setup wizard comp/agent Core agent runtime: loop, agent_init, prompt builder, context-compression, responses endpoint provider/openrouter OpenRouter aggregator provider/qwen Qwen / Alibaba Cloud (OAuth) labels Apr 30, 2026
@alt-glitch

Copy link
Copy Markdown
Collaborator

Implements #6601. Related closed PR: #8481.

teknium1 pushed a commit that referenced this pull request May 17, 2026
qwen3.6-plus did not have an explicit entry in DEFAULT_CONTEXT_LENGTHS,
so the longest-substring fallback matched the generic 'qwen': 131072
catch-all. That dropped the effective context limit from 1,048,576
tokens to 131,072, prematurely lowered the compression threshold, and
produced misleading warnings about main/compression context mismatch
in long sessions.

Add an explicit 'qwen3.6-plus': 1048576 entry before the catch-all and
cover it with a regression test (bare, qwen/, and dashscope/ prefixes).

Note: PR #6599 also mentions touching model_metadata.py but the actual
diff only edits hermes_cli/models.py, so this fix is independent and
not duplicated by that PR.

Closes #27008
gweeteve pushed a commit to gweeteve/hermes-agent that referenced this pull request Jun 2, 2026
qwen3.6-plus did not have an explicit entry in DEFAULT_CONTEXT_LENGTHS,
so the longest-substring fallback matched the generic 'qwen': 131072
catch-all. That dropped the effective context limit from 1,048,576
tokens to 131,072, prematurely lowered the compression threshold, and
produced misleading warnings about main/compression context mismatch
in long sessions.

Add an explicit 'qwen3.6-plus': 1048576 entry before the catch-all and
cover it with a regression test (bare, qwen/, and dashscope/ prefixes).

Note: PR NousResearch#6599 also mentions touching model_metadata.py but the actual
diff only edits hermes_cli/models.py, so this fix is independent and
not duplicated by that PR.

Closes NousResearch#27008
Seven74AI pushed a commit to Seven74AI/hermes-agent that referenced this pull request Jun 13, 2026
qwen3.6-plus did not have an explicit entry in DEFAULT_CONTEXT_LENGTHS,
so the longest-substring fallback matched the generic 'qwen': 131072
catch-all. That dropped the effective context limit from 1,048,576
tokens to 131,072, prematurely lowered the compression threshold, and
produced misleading warnings about main/compression context mismatch
in long sessions.

Add an explicit 'qwen3.6-plus': 1048576 entry before the catch-all and
cover it with a regression test (bare, qwen/, and dashscope/ prefixes).

Note: PR NousResearch#6599 also mentions touching model_metadata.py but the actual
diff only edits hermes_cli/models.py, so this fix is independent and
not duplicated by that PR.

Closes NousResearch#27008
alt-glitch pushed a commit that referenced this pull request Jun 14, 2026
qwen3.6-plus did not have an explicit entry in DEFAULT_CONTEXT_LENGTHS,
so the longest-substring fallback matched the generic 'qwen': 131072
catch-all. That dropped the effective context limit from 1,048,576
tokens to 131,072, prematurely lowered the compression threshold, and
produced misleading warnings about main/compression context mismatch
in long sessions.

Add an explicit 'qwen3.6-plus': 1048576 entry before the catch-all and
cover it with a regression test (bare, qwen/, and dashscope/ prefixes).

Note: PR #6599 also mentions touching model_metadata.py but the actual
diff only edits hermes_cli/models.py, so this fix is independent and
not duplicated by that PR.

Closes #27008
T02200059 pushed a commit to T02200059/hermes-agent that referenced this pull request Jun 18, 2026
qwen3.6-plus did not have an explicit entry in DEFAULT_CONTEXT_LENGTHS,
so the longest-substring fallback matched the generic 'qwen': 131072
catch-all. That dropped the effective context limit from 1,048,576
tokens to 131,072, prematurely lowered the compression threshold, and
produced misleading warnings about main/compression context mismatch
in long sessions.

Add an explicit 'qwen3.6-plus': 1048576 entry before the catch-all and
cover it with a regression test (bare, qwen/, and dashscope/ prefixes).

Note: PR NousResearch#6599 also mentions touching model_metadata.py but the actual
diff only edits hermes_cli/models.py, so this fix is independent and
not duplicated by that PR.

Closes NousResearch#27008
liuchanchen pushed a commit to liuchanchen/hermes-agent that referenced this pull request Jun 23, 2026
qwen3.6-plus did not have an explicit entry in DEFAULT_CONTEXT_LENGTHS,
so the longest-substring fallback matched the generic 'qwen': 131072
catch-all. That dropped the effective context limit from 1,048,576
tokens to 131,072, prematurely lowered the compression threshold, and
produced misleading warnings about main/compression context mismatch
in long sessions.

Add an explicit 'qwen3.6-plus': 1048576 entry before the catch-all and
cover it with a regression test (bare, qwen/, and dashscope/ prefixes).

Note: PR NousResearch#6599 also mentions touching model_metadata.py but the actual
diff only edits hermes_cli/models.py, so this fix is independent and
not duplicated by that PR.

Closes NousResearch#27008
@teknium1

Copy link
Copy Markdown
Contributor

Thanks for the catalog update. The KiloCode entry remains absent from current main, but the PR description no longer matches the patch or current catalog direction.

Problems

  • gh pr diff 6599 shows only the KiloCode list addition. The described OpenRouter and metadata changes are not in the PR; metadata is already implemented at agent/model_metadata.py:277 by d9abbe7fa.
  • Current main deliberately excludes qwen/qwen3.6-plus from the OpenRouter fallback snapshot (hermes_cli/models.py:61-65): ccd3d04fc replaced it with qwen/qwen3.7-max. Reintroducing it needs a current curation decision.

Suggested changes

  • If KiloCode still supports the model, salvage this as a focused KiloCode-catalog update.
  • If OpenRouter should also regain the entry, update the current snapshot and regenerate website/static/api/model-catalog.json; tests/hermes_cli/test_model_catalog.py:514-543 enforces that manifest invariant.

Automated hermes-sweeper review.

@teknium1 teknium1 added sweeper:risk-compatibility Sweeper risk: may break existing users, config, migrations, defaults, or upgrades sweeper:blast-moderate Sweeper blast radius: moderate — a subsystem or single platform labels Jul 12, 2026
@alt-glitch alt-glitch added provider/kilo Kilo Code and removed comp/agent Core agent runtime: loop, agent_init, prompt builder, context-compression, responses endpoint provider/openrouter OpenRouter aggregator sweeper:risk-compatibility Sweeper risk: may break existing users, config, migrations, defaults, or upgrades labels Jul 12, 2026
donbowman pushed a commit to donbowman/hermes-agent that referenced this pull request Jul 13, 2026
qwen3.6-plus did not have an explicit entry in DEFAULT_CONTEXT_LENGTHS,
so the longest-substring fallback matched the generic 'qwen': 131072
catch-all. That dropped the effective context limit from 1,048,576
tokens to 131,072, prematurely lowered the compression threshold, and
produced misleading warnings about main/compression context mismatch
in long sessions.

Add an explicit 'qwen3.6-plus': 1048576 entry before the catch-all and
cover it with a regression test (bare, qwen/, and dashscope/ prefixes).

Note: PR NousResearch#6599 also mentions touching model_metadata.py but the actual
diff only edits hermes_cli/models.py, so this fix is independent and
not duplicated by that PR.

Closes NousResearch#27008
Gravezzz pushed a commit to Gravezzz/hermes-agent that referenced this pull request Jul 21, 2026
qwen3.6-plus did not have an explicit entry in DEFAULT_CONTEXT_LENGTHS,
so the longest-substring fallback matched the generic 'qwen': 131072
catch-all. That dropped the effective context limit from 1,048,576
tokens to 131,072, prematurely lowered the compression threshold, and
produced misleading warnings about main/compression context mismatch
in long sessions.

Add an explicit 'qwen3.6-plus': 1048576 entry before the catch-all and
cover it with a regression test (bare, qwen/, and dashscope/ prefixes).

Note: PR NousResearch#6599 also mentions touching model_metadata.py but the actual
diff only edits hermes_cli/models.py, so this fix is independent and
not duplicated by that PR.

Closes NousResearch#27008

@GottZ GottZ left a comment

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This was generated by AI during triage.

Summary

Four PRs touch the Qwen 3.6 Plus catalog family, but their diffs cover distinct variants or providers: #4376 adds the free variant to OpenRouter and Nous, #6599 and #8481 identically add the paid model to KiloCode, and #14673 adds it to Alibaba. Only #6599/#8481 directly overlap; the other changes are reference implementations or provider-specific fixes already present on main.

Related pull requests

  • #4376 [merged] related — (+2/-0) — merged reference implementation: Adds qwen/qwen3.6-plus-preview:free to the OpenRouter and Nous lists, so it is relevant catalog precedent but does not provide the paid KiloCode entry.
  • #6599 related — (+1/-0) — keep open and salvage: The actual diff only adds qwen/qwen3.6-plus to KiloCode; it does not contain the OpenRouter or metadata changes claimed by the description. Consistent with the keep_open review on #6599, narrow the PR description and merge only after confirming current KiloCode support.
  • #8481 [closed] duplicate — (+1/-0) — duplicate of #6599: Its diff is identical to #6599's KiloCode addition. The earlier closure rationale cited the paid OpenRouter entry already being on main, but that does not match this PR's actual KiloCode-only diff; it should nevertheless remain closed because #6599 is the active duplicate.
  • #14673 [closed] related — (+1/-0) — already fixed on main: Adds qwen3.6-plus to the Alibaba catalog, a separate provider-specific change that the contributor triage confirmed is now present on main, so the closed PR remains useful only as historical reference.

Duplicates

#6599 and #8481 make the same one-line KiloCode catalog change; #4376 and #14673 target different variants or providers and are not duplicates.

Suggested consolidation

Merge #6599 after narrowing it to the KiloCode-only diff and confirming that KiloCode still supports qwen/qwen3.6-plus; this follows the visible keep_open review rather than reintroducing the model into the deliberately changed OpenRouter fallback. Keep #8481 closed as its exact duplicate, while #4376 remains the merged free-variant reference and #14673 remains closed because its Alibaba change is already on main.

Complex graph

flowchart LR
    classDef open fill:#dbeafe,stroke:#1d4ed8,color:#1e3a8a
    classDef merged fill:#dcfce7,stroke:#15803d,color:#14532d
    classDef closed fill:#e5e7eb,stroke:#6b7280,color:#1f2937
    classDef unverified fill:#f3f4f6,stroke:#9ca3af,color:#374151
    classDef best stroke-width:3px,stroke:#b45309
    classDef target stroke-width:3px,stroke:#4338ca
    subgraph Dup6599 ["PRs duplicating each other"]
        P6599["PR #6599 (open)"]
        P8481["PR #8481 (closed)"]
    end
    class P6599 open
    class P8481 closed
    class P6599 target
    click P6599 "https://github.com/NousResearch/hermes-agent/pull/6599"
    click P8481 "https://github.com/NousResearch/hermes-agent/pull/8481"
Loading

Graph: solid arrow = fixes / best fix, dashed arrow = partial or unverified (see edge label); boxed group = PRs duplicating each other; amber border = best fix; indigo border = target; gray node = closed or no verify verdict yet (state tag in the node label).

Cross-PR triage: Reviewed 4 pull requests and 0 issues in this complex. Each diff was read against this issue; Assessment working set: 2 kB of PR diffs, 1 kB of issue/PR text, 2 kB of discussion (6 comments), 1 verify verdict. verdicts reflect diff content, not PR titles. Part of an automated triage batch.

@alt-glitch alt-glitch added needs-repro Bug needs reproduction steps needs-decision Awaiting maintainer decision before any implementation and removed needs-repro Bug needs reproduction steps labels Jul 27, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

comp/cli CLI entry point, hermes_cli/, setup wizard needs-decision Awaiting maintainer decision before any implementation P3 Low — cosmetic, nice to have provider/kilo Kilo Code provider/qwen Qwen / Alibaba Cloud (OAuth) sweeper:blast-moderate Sweeper blast radius: moderate — a subsystem or single platform type/feature New feature or request

Projects

None yet

Development

Successfully merging this pull request may close these issues.

4 participants