HarnessRouter Community Edition: the self-hosted, Apache-2.0 edition of the unified interface for agent harnesses. Run Codex, Claude Code, Hermes, PI, DSH, and more through one API, with sessions, streaming, files, cancellation, and failure handling. Implements the Unified Harness Protocol (UHP), an open standard. Your keys, your infrastructure.
Python agent-apiagent-frameworkagent-harnessai-agentsai-infrastructurecoding-agentconformance-testingdeepseek-harnessharness-engineeringhermesllmopen-protocolopen-sourceopen-standardpi-coding-agentself-hosteduhpunified-apiunified-harness-protocolunified-interface
fast-agent no longer offers gemini-4-argon, nor docker-agent grok-4.20 (deck conversations that changed models) (#607)
github-actions[bot]b2091142cdca
diff --git a/docs/support-matrix-notes.md b/docs/support-matrix-notes.mdindex d3735a3e..07ac92a8 100644--- a/docs/support-matrix-notes.md+++ b/docs/support-matrix-notes.md@@ -5407,3 +5407,20 @@ read the skill. docker-agent expands ``!`command` `` in every local SKILL.md, asskill's line "with morph `!!` prefix awareness" holds that sequence, so the text up to the next backtick ran as acommand, failed, and left an error inside the skill text the model read. The line is reworded in the skillsrepository (HarnessRouter/skills 1553417); an image built from now on carries it.++### Two pairs off after a deck conversation that changed models (0.37.8)++On the hosted service, two pairs failed in deck conversations that changed models along the way, in a way the+matrix's five scenarios do not reach (2026-10-27):++- fast-agent with gemini-4-argon, on Google's own API, after another model's turn (gpt-6.1-sol or+ claude-haiku-4.5) whose history holds the image attach_media added: one tool call and then an empty answer, or+ a promise ("I'll update hello_world.pptx") and no tool call, twice each. argon after argon passed with the+ images, and argon after gpt-6.1-sol passed without the image. On Vercel's route the same history passes.+- docker-agent with grok-4.20, asked to restyle a deck gpt-6.1-sol had made in the same conversation: it said+ the deck was red and made no tool call. On the test instance (Vercel, 0.37.7) the same two turns made the+ change with a real tool call, 2 of 2, and the deck's XML was red.++Both pairs are off the hosted service, and off here alike (`_NOT_OFFERED`, the console's lists, and their rows+in support-matrix-results.json), on Richard's word: "Hide those two pairs". fast-agent keeps the Gemini family+and docker-agent the rest of grok's.diff --git a/docs/support-matrix-results.json b/docs/support-matrix-results.jsonindex 41104411..ec3fe63c 100644--- a/docs/support-matrix-results.json+++ b/docs/support-matrix-results.json@@ -163343,97 +163343,6 @@},"deleted": 200},- "harnessrouter|docker-agent|grok-4.20": {- "provider": "harnessrouter",- "harness": "docker-agent",- "model": "grok-4.20",- "at": "2026-10-25T18:28:19.030Z",- "first": {- "ok": true,- "status": "done",- "pill": "done",- "s": 9.86,- "tail": "Reply with exactly: M1-grok-4.20 M1-grok-4.20 Docker Agent \u00b7 grok-4.20 \u00b7 Done in 7.0 s",- "answer": "M1-grok-4.20",- "turn_files": [],- "why": ""- },- "sid": "hsess999b82f87dba443cb377cc8959ca4802",- "followup": {- "ok": true,- "status": "completed",- "pill": "done",- "s": 8.04,- "tail": "Reply with exactly: M2-grok-4.20 M2-grok-4.20 Docker Agent \u00b7 grok-4.20 \u00b7 Done in 5.3 s",- "answer": "M2-grok-4.20",- "turn_files": [],- "why": ""- },- "switch": {- "to": "gpt-6-sol",- "ok": true,- "status": "completed",- "pill": "done",- "s": 16.51,- "tail": "Reply with exactly: M3-gpt-6-sol M3-gpt-6-sol Docker Agent \u00b7 gpt-6-sol \u00b7 Done in 3.4 s",- "answer": "M3-gpt-6-sol",- "turn_files": [],- "why": ""- },- "artifact": {- "ok": true,- "status": "completed",- "pill": "done",- "s": 7.88,- "tail": "Create a file named hello-docker-agent.txt containing exactly the word HELLO, then reply DONE. Created hello-docker-agent.txt +1 \u22120 DONE txt hello-docker-agent.txt TXT Docker Agent \u00b7 grok-4.20 \u00b7 Done in 5.7 s",- "answer": "DONE",- "turn_files": [- "hello-docker-agent.txt"- ],- "why": "",- "files": [- "hello-docker-agent.txt"- ],- "record_files": [- "hello-docker-agent.txt"- ],- "cards_agree": true- },- "recycle": {- "ok": true,- "status": "completed",- "pill": "done",- "s": 11.25,- "tail": "What exact word did I ask you to reply with in my very first message of this task? Reply with just that word. M1-grok-4.20 Docker Agent \u00b7 grok-4.20 \u00b7 Done in 8.7 s",- "answer": "M1-grok-4.20",- "turn_files": [],- "why": "",- "recycled": {- "code": 200,- "body": "{\"session_id\":\"hsess999b82f87dba443cb377cc8959ca4802\",\"checkpoint_sha\":\"7c5c5588be0c428a4ee8298ec95371570a1c4a4b778d78e8c8146879d74f7711\",\"hydrated\":true,\"hydrate\":{\"ok\":true,\"restored\":true,\"bytes_in"- }- },- "connection": "integration:HarnessRouter API test 1",- "connections": [- "integration:HarnessRouter API test 1"- ],- "served": [- "spacexai/grok-4.20-reasoning",- "openai/gpt-6-sol"- ],- "substituted": [],- "switch_served": [- "openai/gpt-6-sol"- ],- "foreign": [],- "unlabelled": 0,- "tokens": {- "input": 5472,- "cache_read": 9984,- "turns_with_usage": 5- },- "deleted": 200- },"harnessrouter|docker-agent|muse-wonder-1.0": {"provider": "harnessrouter","harness": "docker-agent",@@ -168075,97 +167984,6 @@},"deleted": 200},- "openrouter|docker-agent|grok-4.20": {- "provider": "openrouter",- "harness": "docker-agent",- "model": "grok-4.20",- "at": "2026-10-25T03:46:02.402Z",- "first": {- "ok": true,- "status": "completed",- "pill": "done",- "s": 7.89,- "tail": "Reply with exactly: M1-grok-4.20 M1-grok-4.20 Docker Agent \u00b7 grok-4.20 \u00b7 Done in 3.6 s",- "answer": "M1-grok-4.20",- "turn_files": [],- "why": ""- },- "sid": "hsess288a7d1401844431850538ee8e15963d",- "followup": {- "ok": true,- "status": "completed",- "pill": "done",- "s": 16.19,- "tail": "Reply with exactly: M2-grok-4.20 M2-grok-4.20 Docker Agent \u00b7 grok-4.20 \u00b7 Done in 2.8 s",- "answer": "M2-grok-4.20",- "turn_files": [],- "why": ""- },- "switch": {- "to": "gpt-6.1-sol",- "ok": true,- "status": "completed",- "pill": "done",- "s": 16.16,- "tail": "Reply with exactly: M3-gpt-6.1-sol M3-gpt-6.1-sol Docker Agent \u00b7 gpt-6.1-sol \u00b7 Done in 4.0 s",- "answer": "M3-gpt-6.1-sol",- "turn_files": [],- "why": ""- },- "artifact": {- "ok": true,- "status": "completed",- "pill": "done",- "s": 7.99,- "tail": "Create a file named hello-docker-agent.txt containing exactly the word HELLO, then reply DONE. Created hello-docker-agent.txt +1 \u22120 DONE txt hello-docker-agent.txt TXT Docker Agent \u00b7 grok-4.20 \u00b7 Done in 3.9 s",- "answer": "DONE",- "turn_files": [- "hello-docker-agent.txt"- ],- "why": "",- "files": [- "hello-docker-agent.txt"- ],- "record_files": [- "hello-docker-agent.txt"- ],- "cards_agree": true- },- "recycle": {- "ok": true,- "status": "completed",- "pill": "done",- "s": 16.25,- "tail": "What exact word did I ask you to reply with in my very first message of this task? Reply with just that word. M1-grok-4.20 Docker Agent \u00b7 grok-4.20 \u00b7 Done in 2.7 s",- "answer": "M1-grok-4.20",- "turn_files": [],- "why": "",- "recycled": {- "code": 200,- "body": "{\"session_id\":\"hsess288a7d1401844431850538ee8e15963d\",\"checkpoint_sha\":\"cdfa3c0d1977ccd201af499c57ed1eacc3e8a43b39e6193f6d79de9689e074d0\",\"hydrated\":true,\"hydrate\":{\"ok\":true,\"restored\":true,\"bytes_in"- }- },- "connection": "integration:OpenRouter",- "connections": [- "integration:OpenRouter"- ],- "served": [- "x-ai/grok-4.20",- "openai/gpt-6.1-sol"- ],- "substituted": [],- "switch_served": [- "openai/gpt-6.1-sol"- ],- "foreign": [],- "unlabelled": 0,- "tokens": {- "input": 3021,- "cache_read": 10112,- "turns_with_usage": 5- },- "deleted": 200- },"openrouter|docker-agent|muse-wonder-1.0": {"provider": "openrouter","harness": "docker-agent",@@ -172625,97 +172443,6 @@},"deleted": 200},- "tokenrouter|docker-agent|grok-4.20": {- "provider": "tokenrouter",- "harness": "docker-agent",- "model": "grok-4.20",- "at": "2026-10-25T05:25:45.950Z",- "first": {- "ok": true,- "status": "completed",- "pill": "done",- "s": 7.75,- "tail": "Reply with exactly: M1-grok-4.20 M1-grok-4.20 Docker Agent \u00b7 grok-4.20 \u00b7 Done in 5.8 s",- "answer": "M1-grok-4.20",- "turn_files": [],- "why": ""- },- "sid": "hsess6833ffd556754fdab3ea8bfb98189401",- "followup": {- "ok": true,- "status": "completed",- "pill": "done",- "s": 8.22,- "tail": "Reply with exactly: M2-grok-4.20 M2-grok-4.20 Docker Agent \u00b7 grok-4.20 \u00b7 Done in 4.9 s",- "answer": "M2-grok-4.20",- "turn_files": [],- "why": ""- },- "switch": {- "to": "gpt-6-sol",- "ok": true,- "status": "completed",- "pill": "done",- "s": 7.72,- "tail": "Reply with exactly: M3-gpt-6-sol M3-gpt-6-sol Docker Agent \u00b7 gpt-6-sol \u00b7 Done in 4.8 s",- "answer": "M3-gpt-6-sol",- "turn_files": [],- "why": ""- },- "artifact": {- "ok": true,- "status": "completed",- "pill": "done",- "s": 11.09,- "tail": "Create a file named hello-docker-agent.txt containing exactly the word HELLO, then reply DONE. Created hello-docker-agent.txt +1 \u22120 DONE txt hello-docker-agent.txt TXT Docker Agent \u00b7 grok-4.20 \u00b7 Done in 6.7 s",- "answer": "DONE",- "turn_files": [- "hello-docker-agent.txt"- ],- "why": "",- "files": [- "hello-docker-agent.txt"- ],- "record_files": [- "hello-docker-agent.txt"- ],- "cards_agree": true- },- "recycle": {- "ok": true,- "status": "completed",- "pill": "done",- "s": 7.81,- "tail": "What exact word did I ask you to reply with in my very first message of this task? Reply with just that word. M1-grok-4.20 Docker Agent \u00b7 grok-4.20 \u00b7 Done in 3.7 s",- "answer": "M1-grok-4.20",- "turn_files": [],- "why": "",- "recycled": {- "code": 200,- "body": "{\"session_id\":\"hsess6833ffd556754fdab3ea8bfb98189401\",\"checkpoint_sha\":\"714aeaaf05f53a63c38bb648098895005c868f644d97f767ba84f9eacddb74c8\",\"hydrated\":true,\"hydrate\":{\"ok\":true,\"restored\":true,\"bytes_in"- }- },- "connection": "integration:My TokenRouter",- "connections": [- "integration:My TokenRouter"- ],- "served": [- "x-ai/grok-4.20",- "gpt-6-sol"- ],- "substituted": [],- "switch_served": [- "gpt-6-sol"- ],- "foreign": [],- "unlabelled": 0,- "tokens": {- "input": 9219,- "cache_read": 6912,- "turns_with_usage": 5- },- "deleted": 200- },"tokenrouter|docker-agent|deepseek-v4.1-flash": {"provider": "tokenrouter","harness": "docker-agent",@@ -176538,97 +176265,6 @@},"deleted": 200},- "vercel|docker-agent|grok-4.20": {- "provider": "vercel",- "harness": "docker-agent",- "model": "grok-4.20",- "at": "2026-10-24T23:58:50.354Z",- "first": {- "ok": true,- "status": "done",- "pill": "done",- "s": 10.12,- "tail": "Reply with exactly: M1-grok-4.20 M1-grok-4.20 Docker Agent \u00b7 grok-4.20 \u00b7 Done in 7.1 s",- "answer": "M1-grok-4.20",- "turn_files": [],- "why": ""- },- "sid": "hsessb96a55abb5b54a3a9f83bcef05eacfa2",- "followup": {- "ok": true,- "status": "completed",- "pill": "done",- "s": 8.18,- "tail": "Reply with exactly: M2-grok-4.20 M2-grok-4.20 Docker Agent \u00b7 grok-4.20 \u00b7 Done in 4.3 s",- "answer": "M2-grok-4.20",- "turn_files": [],- "why": ""- },- "switch": {- "to": "gpt-6.1-sol",- "ok": true,- "status": "completed",- "pill": "done",- "s": 8.45,- "tail": "Reply with exactly: M3-gpt-6.1-sol M3-gpt-6.1-sol Docker Agent \u00b7 gpt-6.1-sol \u00b7 Done in 5.2 s",- "answer": "M3-gpt-6.1-sol",- "turn_files": [],- "why": ""- },- "artifact": {- "ok": true,- "status": "completed",- "pill": "done",- "s": 8.08,- "tail": "Create a file named hello-docker-agent.txt containing exactly the word HELLO, then reply DONE. Created hello-docker-agent.txt +1 \u22120 DONE txt hello-docker-agent.txt TXT Docker Agent \u00b7 grok-4.20 \u00b7 Done in 4.9 s",- "answer": "DONE",- "turn_files": [- "hello-docker-agent.txt"- ],- "why": "",- "files": [- "hello-docker-agent.txt"- ],- "record_files": [- "hello-docker-agent.txt"- ],- "cards_agree": true- },- "recycle": {- "ok": true,- "status": "done",- "pill": "done",- "s": 10.42,- "tail": "What exact word did I ask you to reply with in my very first message of this task? Reply with just that word. M1-grok-4.20 Docker Agent \u00b7 grok-4.20 \u00b7 Done in 7.2 s",- "answer": "M1-grok-4.20",- "turn_files": [],- "why": "",- "recycled": {- "code": 200,- "body": "{\"session_id\":\"hsessb96a55abb5b54a3a9f83bcef05eacfa2\",\"checkpoint_sha\":\"07aa7a8fc494aa4e4aae14e297f42e6daa73624fe619319e6609f7d3b5e12677\",\"hydrated\":true,\"hydrate\":{\"ok\":true,\"restored\":true,\"bytes_in"- }- },- "connection": "integration:Vercel AI Gateway",- "connections": [- "integration:Vercel AI Gateway"- ],- "served": [- "spacexai/grok-4.20-reasoning",- "openai/gpt-6.1-sol"- ],- "substituted": [],- "switch_served": [- "openai/gpt-6.1-sol"- ],- "foreign": [],- "unlabelled": 0,- "tokens": {- "input": 8611,- "cache_read": 9984,- "turns_with_usage": 5- },- "deleted": 200- },"vercel|docker-agent|muse-wonder-1.0": {"provider": "vercel","harness": "docker-agent",@@ -181341,96 +180977,10 @@"to": "gpt-5.4-mini","ok": true,"status": "completed",- "pill": "done",- "s": 14.37,- "tail": "Reply with exactly: M3-gpt-5.4-mini M3-gpt-5.4-mini fast-agent \u00b7 gpt-5.4-mini \u00b7 Done in 12.0 s",- "answer": "M3-gpt-5.4-mini",- "turn_files": [],- "why": ""- },- "artifact": {- "ok": true,- "status": "completed",- "pill": "done",- "s": 14.17,- "tail": "Create a file named hello-fast-agent.txt containing exactly the word HELLO, then reply DONE. Ran printf '%s' 'HELLO' > hello-fast-agent.txt DONE txt hello-fast-agent.txt TXT fast-agent \u00b7 gpt-5.2 \u00b7 Done in 11.4 s",- "answer": "DONE",- "turn_files": [- "hello-fast-agent.txt"- ],- "why": "",- "files": [- "hello-fast-agent.txt"- ],- "record_files": [- "hello-fast-agent.txt"- ],- "cards_agree": true- },- "recycle": {- "ok": true,- "status": "done",- "pill": "done",- "s": 12.64,- "tail": "What exact word did I ask you to reply with in my very first message of this task? Reply with just that word. M1-gpt-5.2 fast-agent \u00b7 gpt-5.2 \u00b7 Done in 9.6 s",- "answer": "M1-gpt-5.2",- "turn_files": [],- "why": "",- "recycled": {- "code": 200,- "body": "{\"session_id\":\"hsessed605a79db7f4b39b7583baf665b3b59\",\"checkpoint_sha\":\"68acb6d00b81f1722939e98ea5ec48e7eb59275a63b3637268bfaf9bd12239fa\",\"hydrated\":true,\"hydrate\":{\"ok\":true,\"restored\":true,\"bytes_in"- }- },- "connection": "integration:Custom OpenAI Chat",- "connections": [- "integration:Custom OpenAI Chat"- ],- "served": [- "gpt-5.2-2025-12-11",- "gpt-5.4-mini-2026-03-17"- ],- "substituted": [],- "switch_served": [- "gpt-5.4-mini-2026-03-17"- ],- "foreign": [],- "unlabelled": 0,- "deleted": 200- },- "google|fast-agent|gemini-4-argon": {- "provider": "google",- "harness": "fast-agent",- "model": "gemini-4-argon",- "at": "2026-10-25T06:47:28.372Z",- "first": {- "ok": true,- "status": "completed",- "pill": "done",- "s": 10.76,- "tail": "Reply with exactly: M1-gemini-4-argon M1-gemini-4-argon fast-agent \u00b7 gemini-4-argon \u00b7 Done in 8.8 s",- "answer": "M1-gemini-4-argon",- "turn_files": [],- "why": ""- },- "sid": "hsess85f6dcd1618842d0a7215a0a24b285dd",- "followup": {- "ok": true,- "status": "completed",- "pill": "done",- "s": 19.91,- "tail": "Reply with exactly: M2-gemini-4-argon M2-gemini-4-argon fast-agent \u00b7 gemini-4-argon \u00b7 Done in 17.5 s",- "answer": "M2-gemini-4-argon",- "turn_files": [],- "why": ""- },- "switch": {- "to": "gemini-3.8-flash",- "ok": true,- "status": "completed",- "pill": "done",- "s": 10.74,- "tail": "Reply with exactly: M3-gemini-3.8-flash M3-gemini-3.8-flash fast-agent \u00b7 gemini-3.8-flash \u00b7 Done in 7.6 s",- "answer": "M3-gemini-3.8-flash",+ "pill": "done",+ "s": 14.37,+ "tail": "Reply with exactly: M3-gpt-5.4-mini M3-gpt-5.4-mini fast-agent \u00b7 gpt-5.4-mini \u00b7 Done in 12.0 s",+ "answer": "M3-gpt-5.4-mini","turn_files": [],"why": ""},@@ -181438,8 +180988,8 @@"ok": true,"status": "completed","pill": "done",- "s": 26,- "tail": "Create a file named hello-fast-agent.txt containing exactly the word HELLO, then reply DONE. Edited hello-fast-agent.txt +1 \u22120 DONE txt hello-fast-agent.txt TXT fast-agent \u00b7 gemini-4-argon \u00b7 Done in 21.6 s",+ "s": 14.17,+ "tail": "Create a file named hello-fast-agent.txt containing exactly the word HELLO, then reply DONE. Ran printf '%s' 'HELLO' > hello-fast-agent.txt DONE txt hello-fast-agent.txt TXT fast-agent \u00b7 gpt-5.2 \u00b7 Done in 11.4 s","answer": "DONE","turn_files": ["hello-fast-agent.txt"@@ -181455,29 +181005,29 @@},"recycle": {"ok": true,- "status": "completed",+ "status": "done","pill": "done",- "s": 47.29,- "tail": "What exact word did I ask you to reply with in my very first message of this task? Reply with just that word. M1-gemini-4-argon fast-agent \u00b7 gemini-4-argon \u00b7 Done in 44.9 s",- "answer": "M1-gemini-4-argon",+ "s": 12.64,+ "tail": "What exact word did I ask you to reply with in my very first message of this task? Reply with just that word. M1-gpt-5.2 fast-agent \u00b7 gpt-5.2 \u00b7 Done in 9.6 s",+ "answer": "M1-gpt-5.2","turn_files": [],"why": "","recycled": {"code": 200,- "body": "{\"session_id\":\"hsess85f6dcd1618842d0a7215a0a24b285dd\",\"checkpoint_sha\":\"64093f7b1e635140d4a9836832ced836f0b5d7e13b1bc85d1a8e69fe9179f7d4\",\"hydrated\":true,\"hydrate\":{\"ok\":true,\"restored\":true,\"bytes_in"+ "body": "{\"session_id\":\"hsessed605a79db7f4b39b7583baf665b3b59\",\"checkpoint_sha\":\"68acb6d00b81f1722939e98ea5ec48e7eb59275a63b3637268bfaf9bd12239fa\",\"hydrated\":true,\"hydrate\":{\"ok\":true,\"restored\":true,\"bytes_in"}},- "connection": "integration:Google AI Studio",+ "connection": "integration:Custom OpenAI Chat","connections": [- "integration:Google AI Studio"+ "integration:Custom OpenAI Chat"],"served": [- "gemini-4-argon",- "gemini-3.8-flash"+ "gpt-5.2-2025-12-11",+ "gpt-5.4-mini-2026-03-17"],"substituted": [],"switch_served": [- "gemini-3.8-flash"+ "gpt-5.4-mini-2026-03-17"],"foreign": [],"unlabelled": 0,@@ -183905,97 +183455,6 @@},"deleted": 200},- "harnessrouter|fast-agent|gemini-4-argon": {- "provider": "harnessrouter",- "harness": "fast-agent",- "model": "gemini-4-argon",- "at": "2026-10-25T18:37:24.241Z",- "first": {- "ok": true,- "status": "completed",- "pill": "done",- "s": 13.84,- "tail": "Reply with exactly: M1-gemini-4-argon M1-gemini-4-argon fast-agent \u00b7 gemini-4-argon \u00b7 Done in 10.8 s",- "answer": "M1-gemini-4-argon",- "turn_files": [],- "why": ""- },- "sid": "hsess9e346dfe0105472980e37e2c71ea8c75",- "followup": {- "ok": true,- "status": "completed",- "pill": "done",- "s": 13.83,- "tail": "Reply with exactly: M2-gemini-4-argon M2-gemini-4-argon fast-agent \u00b7 gemini-4-argon \u00b7 Done in 10.9 s",- "answer": "M2-gemini-4-argon",- "turn_files": [],- "why": ""- },- "switch": {- "to": "gpt-6-sol",- "ok": true,- "status": "completed",- "pill": "done",- "s": 13.82,- "tail": "Reply with exactly: M3-gpt-6-sol M3-gpt-6-sol fast-agent \u00b7 gpt-6-sol \u00b7 Done in 11.8 s",- "answer": "M3-gpt-6-sol",- "turn_files": [],- "why": ""- },- "artifact": {- "ok": true,- "status": "completed",- "pill": "done",- "s": 26.67,- "tail": "Create a file named hello-fast-agent.txt containing exactly the word HELLO, then reply DONE. Used Write text file DONE txt hello-fast-agent.txt TXT fast-agent \u00b7 gemini-4-argon \u00b7 Done in 21.8 s",- "answer": "DONE",- "turn_files": [- "hello-fast-agent.txt"- ],- "why": "",- "files": [- "hello-fast-agent.txt"- ],- "record_files": [- "hello-fast-agent.txt"- ],- "cards_agree": true- },- "recycle": {- "ok": true,- "status": "completed",- "pill": "done",- "s": 20.02,- "tail": "What exact word did I ask you to reply with in my very first message of this task? Reply with just that word. M1-gemini-4-argon fast-agent \u00b7 gemini-4-argon \u00b7 Done in 17.1 s",- "answer": "M1-gemini-4-argon",- "turn_files": [],- "why": "",- "recycled": {- "code": 200,- "body": "{\"session_id\":\"hsess9e346dfe0105472980e37e2c71ea8c75\",\"checkpoint_sha\":\"e2ce7310e6b01bcc1603a49cc54aa397b2d56409447edebd772cd2436c142527\",\"hydrated\":true,\"hydrate\":{\"ok\":true,\"restored\":true,\"bytes_in"- }- },- "connection": "integration:HarnessRouter API test 1",- "connections": [- "integration:HarnessRouter API test 1"- ],- "served": [- "gemini-4-argon",- "openai/gpt-6-sol"- ],- "substituted": [],- "switch_served": [- "openai/gpt-6-sol"- ],- "foreign": [],- "unlabelled": 0,- "tokens": {- "input": 5631,- "cache_read": 0,- "turns_with_usage": 5- },- "deleted": 200- },"harnessrouter|fast-agent|gemini-3.8-flash": {"provider": "harnessrouter","harness": "fast-agent",@@ -189898,92 +189357,6 @@"unlabelled": 0,"deleted": 200},- "openrouter|fast-agent|gemini-4-argon": {- "provider": "openrouter",- "harness": "fast-agent",- "model": "gemini-4-argon",- "at": "2026-10-25T09:49:53.083Z",- "first": {- "ok": true,- "status": "completed",- "pill": "done",- "s": 36.43,- "tail": "Reply with exactly: M1-gemini-4-argon M1-gemini-4-argon fast-agent \u00b7 gemini-4-argon \u00b7 Done in 34.3 s",- "answer": "M1-gemini-4-argon",- "turn_files": [],- "why": ""- },- "sid": "hsess36ef568ed18f41d99fb05e7bd5f4e206",- "followup": {- "ok": true,- "status": "completed",- "pill": "done",- "s": 20.67,- "tail": "Reply with exactly: M2-gemini-4-argon M2-gemini-4-argon fast-agent \u00b7 gemini-4-argon \u00b7 Done in 16.7 s",- "answer": "M2-gemini-4-argon",- "turn_files": [],- "why": ""- },- "switch": {- "to": "gpt-6.1-sol",- "ok": true,- "status": "done",- "pill": "done",- "s": 19.46,- "tail": "Reply with exactly: M3-gpt-6.1-sol M3-gpt-6.1-sol fast-agent \u00b7 gpt-6.1-sol \u00b7 Done in 16.7 s",- "answer": "M3-gpt-6.1-sol",- "turn_files": [],- "why": ""- },- "artifact": {- "ok": true,- "status": "completed",- "pill": "done",- "s": 23.57,- "tail": "Create a file named hello-fast-agent.txt containing exactly the word HELLO, then reply DONE. Edited hello-fast-agent.txt +1 \u22120 DONE txt hello-fast-agent.txt TXT fast-agent \u00b7 gemini-4-argon \u00b7 Done in 20.7 s",- "answer": "DONE",- "turn_files": [- "hello-fast-agent.txt"- ],- "why": "",- "files": [- "hello-fast-agent.txt"- ],- "record_files": [- "hello-fast-agent.txt"- ],- "cards_agree": true- },- "recycle": {- "ok": true,- "status": "done",- "pill": "done",- "s": 19.94,- "tail": "What exact word did I ask you to reply with in my very first message of this task? Reply with just that word. M1-gemini-4-argon fast-agent \u00b7 gemini-4-argon \u00b7 Done in 17.1 s",- "answer": "M1-gemini-4-argon",- "turn_files": [],- "why": "",- "recycled": {- "code": 200,- "body": "{\"session_id\":\"hsess36ef568ed18f41d99fb05e7bd5f4e206\",\"checkpoint_sha\":\"6bc0fbf92155719da242fc2950921fd4aac5a409d9c59f5f5fa072a15f113af0\",\"hydrated\":true,\"hydrate\":{\"ok\":true,\"restored\":true,\"bytes_in"- }- },- "connection": "integration:OpenRouter",- "connections": [- "integration:OpenRouter"- ],- "served": [- "google/gemini-4-argon",- "openai/gpt-6.1-sol"- ],- "substituted": [],- "switch_served": [- "openai/gpt-6.1-sol"- ],- "foreign": [],- "unlabelled": 0,- "deleted": 200- },"openrouter|fast-agent|gemini-3.8-flash": {"provider": "openrouter","harness": "fast-agent",@@ -196023,92 +195396,6 @@"unlabelled": 0,"deleted": 200},- "vercel|fast-agent|gemini-4-argon": {- "provider": "vercel",- "harness": "fast-agent",- "model": "gemini-4-argon",- "at": "2026-10-24T23:42:47.629Z",- "first": {- "ok": true,- "status": "completed",- "pill": "done",- "s": 20.37,- "tail": "Reply with exactly: M1-gemini-4-argon M1-gemini-4-argon fast-agent \u00b7 gemini-4-argon \u00b7 Done in 17.9 s",- "answer": "M1-gemini-4-argon",- "turn_files": [],- "why": ""- },- "sid": "hsess838540b7fa244c3ab734281c96768d4f",- "followup": {- "ok": true,- "status": "completed",- "pill": "done",- "s": 17.5,- "tail": "Reply with exactly: M2-gemini-4-argon M2-gemini-4-argon fast-agent \u00b7 gemini-4-argon \u00b7 Done in 14.2 s",- "answer": "M2-gemini-4-argon",- "turn_files": [],- "why": ""- },- "switch": {- "to": "gpt-6.1-sol",- "ok": true,- "status": "completed",- "pill": "done",- "s": 17.31,- "tail": "Reply with exactly: M3-gpt-6.1-sol M3-gpt-6.1-sol fast-agent \u00b7 gpt-6.1-sol \u00b7 Done in 14.2 s",- "answer": "M3-gpt-6.1-sol",- "turn_files": [],- "why": ""- },- "artifact": {- "ok": true,- "status": "completed",- "pill": "done",- "s": 20.46,- "tail": "Create a file named hello-fast-agent.txt containing exactly the word HELLO, then reply DONE. Edited hello-fast-agent.txt +1 \u22120 DONE txt hello-fast-agent.txt TXT fast-agent \u00b7 gemini-4-argon \u00b7 Done in 16.8 s",- "answer": "DONE",- "turn_files": [- "hello-fast-agent.txt"- ],- "why": "",- "files": [- "hello-fast-agent.txt"- ],- "record_files": [- "hello-fast-agent.txt"- ],- "cards_agree": true- },- "recycle": {- "ok": true,- "status": "completed",- "pill": "done",- "s": 14.71,- "tail": "What exact word did I ask you to reply with in my very first message of this task? Reply with just that word. M1-gemini-4-argon fast-agent \u00b7 gemini-4-argon \u00b7 Done in 12.6 s",- "answer": "M1-gemini-4-argon",- "turn_files": [],- "why": "",- "recycled": {- "code": 200,- "body": "{\"session_id\":\"hsess838540b7fa244c3ab734281c96768d4f\",\"checkpoint_sha\":\"7cd35e390d2c65ac77a451faf6d606a20ef9d1f0c09b6d394a1669d198d78cf8\",\"hydrated\":true,\"hydrate\":{\"ok\":true,\"restored\":true,\"bytes_in"- }- },- "connection": "integration:Vercel AI Gateway",- "connections": [- "integration:Vercel AI Gateway"- ],- "served": [- "google/gemini-4-argon",- "openai/gpt-6.1-sol"- ],- "substituted": [],- "switch_served": [- "openai/gpt-6.1-sol"- ],- "foreign": [],- "unlabelled": 0,- "deleted": 200- },"vercel|fast-agent|gemini-3.8-flash": {"provider": "vercel","harness": "fast-agent",diff --git a/docs/support-matrix.md b/docs/support-matrix.mdindex 8c59a2dc..5f44b2ce 100644--- a/docs/support-matrix.md+++ b/docs/support-matrix.md@@ -2434,7 +2434,6 @@ Not run in this column, 331 pairs the provider serves that the harness did not r| fast-agent | gemini-3.5-flash-lite | pass | pass | pass (gemini-4-argon) | pass | pass | Google AI Studio | || fast-agent | gemini-3.6-flash | pass | pass | pass (gemini-4-argon) | pass | pass | Google AI Studio | || fast-agent | gemini-3.8-flash | pass | pass | pass (gemini-4-argon) | pass | pass | Google AI Studio | |-| fast-agent | gemini-4-argon | pass | pass | pass (gemini-3.8-flash) | pass | pass | Google AI Studio | || hermes | gemini-3.6-flash | n/a | n/a | n/a | n/a | n/a | ? | re-run on the sponsored Tier 3 key and 0.13.14 (the relays drop the field Google refuses); first try on the Free-tier ke ; retested once; first try: first Your openai-api key was refused: API call failed after 3 retries: HTTP 429: [{|| opencode | gemini-3.6-flash | pass | pass | n/a | FAIL | pass | Google AI Studio | artifact: [{"connection": "integration:Google AI Studio", "status": "failed", "error": "Bad Request: [{\n \"error\": {\n \"code\": 400,\n \"mes ; re-run on the sponsored Tier 3 key and 0.13.14 (the relays drop the field Google refuses); first try on the Free-tier ke ; retested once; first try: followup [{"connection": "integration:Google AI Studio", "status": "failed", "error": "To; artifact [{"connection": "integration:Google AI Studio", "status": "failed", "error": "To; recycle [{"connection": "integration:Google AI Studio", "status": "failed", "error": "To |@@ -2449,9 +2448,9 @@ Not run in this column, 331 pairs the provider serves that the harness did not r| vibe | gemini-3.8-flash | pass | pass | pass (gemini-4-argon) | pass | pass | Google AI Studio | cache 0% of input || vibe | gemini-4-argon | pass | pass | pass (gemini-3.8-flash) | pass | pass | Google AI Studio | cache 6% of input |-29 pairs, 126 of 128 scenario runs passed.+28 pairs, 121 of 123 scenario runs passed.-Not run in this column, 35 pairs the provider serves that the harness did not run, with the reason:+Not run in this column, 36 pairs the provider serves that the harness did not run, with the reason:- dsh x gemini-3-flash-preview: not run, not run in this column- dsh x gemini-3.1-flash-lite: not run, not run in this column@@ -2460,6 +2459,7 @@ Not run in this column, 35 pairs the provider serves that the harness did not ru- dsh x gemini-3.5-flash-lite: not run, not run in this column- dsh x gemini-3.8-flash: not run, not run in this column- dsh x gemini-4-argon: not run, not run in this column+- fast-agent x gemini-4-argon: not run, not in this harness's catalog (unmeasured or excluded, see the catalog's note)- hermes x gemini-3-flash-preview: not run, not run in this column- hermes x gemini-3.1-flash-lite: not run, not run in this column- hermes x gemini-3.1-pro-preview: not run, not run in this column@@ -2982,7 +2982,6 @@ Findings, pairs served by a connection other than the one under test or as a mod| docker-agent | gpt-5.6-terra | pass | pass | pass (gpt-6-sol) | pass | pass | HarnessRouter API test 1 | cache 85% of input || docker-agent | gpt-6-luna | pass | pass | pass (gpt-6-sol) | pass | pass | HarnessRouter API test 1 | cache 67% of input || docker-agent | gpt-6-sol | pass | pass | pass (gpt-6-luna) | pass | pass | HarnessRouter API test 1 | cache 72% of input |-| docker-agent | grok-4.20 | pass | pass | pass (gpt-6-sol) | pass | pass | HarnessRouter API test 1 | cache 65% of input || docker-agent | grok-4.3 | pass | pass | pass (gpt-6-sol) | pass | pass | HarnessRouter API test 1 | cache 75% of input || docker-agent | grok-4.6 | pass | pass | pass (gpt-6-sol) | pass | pass | HarnessRouter API test 1 | cache 54% of input || docker-agent | hunyuan-4-preview | pass | pass | pass (gpt-6-sol) | pass | pass | HarnessRouter API test 1 | cache 70% of input |@@ -3023,7 +3022,6 @@ Findings, pairs served by a connection other than the one under test or as a mod| fast-agent | gemini-3.6-flash | pass | pass | pass (gpt-6-sol) | pass | pass | HarnessRouter API test 1 | cache 0% of input || fast-agent | gemini-3.7-flash | pass | pass | pass (gpt-6-sol) | pass | pass | HarnessRouter API test 1 | cache 0% of input || fast-agent | gemini-3.8-flash | pass | pass | pass (gpt-6-sol) | pass | pass | HarnessRouter API test 1 | cache 0% of input |-| fast-agent | gemini-4-argon | pass | pass | pass (gpt-6-sol) | pass | pass | HarnessRouter API test 1 | cache 0% of input || fast-agent | glm-5.3 | pass | pass | pass (gpt-6-sol) | pass | pass | HarnessRouter API test 1 | cache 66% of input || fast-agent | glm-5.3-flash | pass | pass | pass (gpt-6-sol) | pass | pass | HarnessRouter API test 1 | cache 0% of input || fast-agent | gpt-5.2 | pass | pass | pass (gpt-6-sol) | pass | pass | HarnessRouter API test 1 | cache 56% of input |@@ -3104,7 +3102,7 @@ Findings, pairs served by a connection other than the one under test or as a mod| vibe | qwen3.8-max | pass | pass | pass (gpt-6-sol) | pass | pass | HarnessRouter API test 1 | cache 51% of input || vibe | step-3.7-flash | pass | pass | pass (gpt-6-sol) | pass | pass | HarnessRouter API test 1 | cache 49% of input |-197 pairs, 985 of 985 scenario runs passed.+195 pairs, 975 of 975 scenario runs passed.Findings, pairs served by a connection other than the one under test or as a model other than the id asked for:@@ -3151,7 +3149,7 @@ Findings, pairs served by a connection other than the one under test or as a mod- vibe x claude-sonnet-5: no prompt cache reads over 77462 input tokens (every call paid full price)- vibe x claude-sonnet-5.5: no prompt cache reads over 77545 input tokens (every call paid full price)-Not run in this column, 34 pairs the provider serves that the harness did not run, with the reason:+Not run in this column, 35 pairs the provider serves that the harness did not run, with the reason:- dcode x gemini-3-flash-preview: not run, this provider's connection refuses the pair for this harness (the gateway's channel rules)- dcode x gemini-3.1-pro-preview: not run, this provider's connection refuses the pair for this harness (the gateway's channel rules)@@ -3169,6 +3167,7 @@ Not run in this column, 34 pairs the provider serves that the harness did not ru- docker-agent x gpt-5.3-codex: not run, the model answers on the Responses API only and this harness speaks chat/completions only- docker-agent x gpt-6-astra: not run, the model answers on the Responses API only and this harness speaks chat/completions only- docker-agent x gpt-6.1-sol: not run, this provider's connection refuses the pair for this harness (the gateway's channel rules)+- docker-agent x grok-4.20: not run, not in this harness's catalog (unmeasured or excluded, see the catalog's note)- docker-agent x grok-4.5: not run, not in this harness's catalog (unmeasured or excluded, see the catalog's note)- docker-agent x grok-build-0.1: not run, not in this harness's catalog (unmeasured or excluded, see the catalog's note)- docker-agent x nemotron-3.5-lightning: not run, not in this harness's catalog (unmeasured or excluded, see the catalog's note)@@ -4002,7 +4001,6 @@ Not run in this column, 61 pairs the provider serves that the harness did not ru| docker-agent | gpt-6-luna | pass | pass | pass (gpt-6.1-sol) | pass | pass | OpenRouter | cache 100% of input || docker-agent | gpt-6-sol | pass | pass | pass (gpt-6.1-sol) | pass | pass | OpenRouter | cache 100% of input || docker-agent | gpt-6.1-sol | pass | pass | pass (gpt-6-sol) | pass | pass | OpenRouter | cache 100% of input |-| docker-agent | grok-4.20 | pass | pass | pass (gpt-6.1-sol) | pass | pass | OpenRouter | cache 77% of input || docker-agent | grok-4.3 | pass | pass | pass (gpt-6.1-sol) | pass | pass | OpenRouter | cache 53% of input || docker-agent | grok-4.6 | pass | pass | pass (gpt-6.1-sol) | pass | pass | OpenRouter | cache 82% of input || docker-agent | hunyuan-4-preview | pass | pass | pass (gpt-6.1-sol) | pass | pass | OpenRouter | cache 80% of input |@@ -4074,7 +4072,6 @@ Not run in this column, 61 pairs the provider serves that the harness did not ru| fast-agent | gemini-3.6-flash | pass | pass | pass (gpt-6.1-sol) | pass | pass | OpenRouter | || fast-agent | gemini-3.7-flash | pass | pass | pass (gpt-6.1-sol) | pass | pass | OpenRouter | || fast-agent | gemini-3.8-flash | pass | pass | pass (gpt-6.1-sol) | pass | pass | OpenRouter | |-| fast-agent | gemini-4-argon | pass | pass | pass (gpt-6.1-sol) | pass | pass | OpenRouter | || fast-agent | glm-5.3 | pass | pass | pass (gpt-6.1-sol) | pass | pass | OpenRouter | || fast-agent | glm-5.3-flash | pass | pass | pass (gpt-6.1-sol) | pass | pass | OpenRouter | || fast-agent | gpt-5.2 | pass | pass | pass (gpt-6.1-sol) | pass | pass | OpenRouter | |@@ -4376,7 +4373,7 @@ session_id: 20260913_060549_70cbeb || vibe | qwen3.8-max | pass | pass | pass (gpt-6.1-sol) | pass | pass | OpenRouter | cache 53% of input || vibe | step-3.7-flash | pass | pass | pass (gpt-6.1-sol) | pass | pass | OpenRouter | cache 65% of input |-491 pairs, 2414 of 2420 scenario runs passed; 7 pairs served by another connection or as another model are findings, not counted.+489 pairs, 2404 of 2410 scenario runs passed; 7 pairs served by another connection or as another model are findings, not counted.Findings, pairs served by a connection other than the one under test or as a model other than the id asked for:@@ -4412,7 +4409,7 @@ Findings, pairs served by a connection other than the one under test or as a mod- pi x claude-haiku-5.5: no prompt cache reads over 25529 input tokens (every call paid full price)- qwen x claude-haiku-5.5: no prompt cache reads over 570679 input tokens (every call paid full price)-Not run in this column, 937 pairs the provider serves that the harness did not run, with the reason:+Not run in this column, 939 pairs the provider serves that the harness did not run, with the reason:- agentzero x claude-fable-5: not run, not in this harness's catalog (unmeasured or excluded, see the catalog's note)- agentzero x claude-fable-5-1: not run, not in this harness's catalog (unmeasured or excluded, see the catalog's note)@@ -4739,6 +4736,7 @@ Not run in this column, 937 pairs the provider serves that the harness did not r- docker-agent x claude-opus-4.7: not run, not in this harness's catalog (unmeasured or excluded, see the catalog's note)- docker-agent x gpt-5.3-codex: not run, the model answers on the Responses API only and this harness speaks chat/completions only- docker-agent x gpt-6-astra: not run, the model answers on the Responses API only and this harness speaks chat/completions only+- docker-agent x grok-4.20: not run, not in this harness's catalog (unmeasured or excluded, see the catalog's note)- docker-agent x grok-4.5: not run, not in this harness's catalog (unmeasured or excluded, see the catalog's note)- docker-agent x grok-build-0.1: not run, not in this harness's catalog (unmeasured or excluded, see the catalog's note)- docker-agent x hunyuan-3: not run, not in this harness's catalog (unmeasured or excluded, see the catalog's note)@@ -4788,6 +4786,7 @@ Not run in this column, 937 pairs the provider serves that the harness did not r- dsh x qwen3.7-plus: not run, not run in this column- dsh x qwen3.8-27b: not run, not run in this column- dsh x qwen3.8-flash: not run, not run in this column+- fast-agent x gemini-4-argon: not run, not in this harness's catalog (unmeasured or excluded, see the catalog's note)- fast-agent x gpt-5.3-codex: not run, the model answers on the Responses API only and this harness speaks chat/completions only- fast-agent x gpt-6-astra: not run, the model answers on the Responses API only and this harness speaks chat/completions only- fast-agent x grok-4.5: not run, not in this harness's catalog (unmeasured or excluded, see the catalog's note)@@ -5441,7 +5440,6 @@ Not run in this column, 937 pairs the provider serves that the harness did not r| docker-agent | gpt-5.6-terra | pass | pass | pass (gpt-6-sol) | pass | pass | My TokenRouter | cache 100% of input || docker-agent | gpt-6-luna | pass | pass | pass (gpt-6-sol) | pass | pass | My TokenRouter | cache 100% of input || docker-agent | gpt-6-sol | pass | pass | pass (gpt-6-luna) | pass | pass | My TokenRouter | cache 100% of input |-| docker-agent | grok-4.20 | pass | pass | pass (gpt-6-sol) | pass | pass | My TokenRouter | cache 43% of input || docker-agent | grok-4.3 | pass | pass | pass (gpt-6-sol) | pass | pass | My TokenRouter | cache 81% of input || docker-agent | grok-4.6 | pass | pass | pass (gpt-6-sol) | pass | pass | My TokenRouter | cache 88% of input || docker-agent | hunyuan-4-preview | pass | pass | pass (gpt-6-sol) | pass | pass | My TokenRouter | cache 38% of input |@@ -5626,7 +5624,7 @@ Not run in this column, 937 pairs the provider serves that the harness did not r| vibe | qwen3.8-max | pass | pass | pass (gpt-6-sol) | pass | pass | My TokenRouter | cache 57% of input || vibe | step-3.7-flash | pass | pass | pass (gpt-6-sol) | pass | pass | My TokenRouter | cache 32% of input |-269 pairs, 1234 of 1236 scenario runs passed; 21 pairs served by another connection or as another model are findings, not counted.+268 pairs, 1229 of 1231 scenario runs passed; 21 pairs served by another connection or as another model are findings, not counted.Findings, pairs served by a connection other than the one under test or as a model other than the id asked for:@@ -5679,7 +5677,7 @@ Findings, pairs served by a connection other than the one under test or as a mod- vibe x claude-sonnet-5: no prompt cache reads over 77995 input tokens (every call paid full price)- vibe x claude-sonnet-5.5: no prompt cache reads over 77418 input tokens (every call paid full price)-Not run in this column, 595 pairs the provider serves that the harness did not run, with the reason:+Not run in this column, 596 pairs the provider serves that the harness did not run, with the reason:- aider x claude-fable-5: not run, not run in this column- aider x claude-fable-5-1: not run, not run in this column@@ -5917,6 +5915,7 @@ Not run in this column, 595 pairs the provider serves that the harness did not r- docker-agent x gpt-5.3-codex: not run, the model answers on the Responses API only and this harness speaks chat/completions only- docker-agent x gpt-6-astra: not run, the model answers on the Responses API only and this harness speaks chat/completions only- docker-agent x gpt-6.1-sol: not run, this provider's connection refuses the pair for this harness (the gateway's channel rules)+- docker-agent x grok-4.20: not run, not in this harness's catalog (unmeasured or excluded, see the catalog's note)- docker-agent x grok-4.5: not run, not in this harness's catalog (unmeasured or excluded, see the catalog's note)- docker-agent x grok-build-0.1: not run, not in this harness's catalog (unmeasured or excluded, see the catalog's note)- docker-agent x nemotron-3.5-lightning: not run, not in this harness's catalog (unmeasured or excluded, see the catalog's note)@@ -6404,7 +6403,6 @@ Not run in this column, 595 pairs the provider serves that the harness did not r| docker-agent | gpt-6-luna | pass | pass | pass (gpt-6.1-sol) | pass | pass | Vercel AI Gateway | cache 67% of input || docker-agent | gpt-6-sol | pass | pass | pass (gpt-6.1-sol) | pass | pass | Vercel AI Gateway | cache 72% of input || docker-agent | gpt-6.1-sol | pass | pass | pass (gpt-6-sol) | pass | pass | Vercel AI Gateway | cache 72% of input |-| docker-agent | grok-4.20 | pass | pass | pass (gpt-6.1-sol) | pass | pass | Vercel AI Gateway | cache 54% of input || docker-agent | grok-4.3 | pass | pass | pass (gpt-6.1-sol) | pass | pass | Vercel AI Gateway | cache 54% of input || docker-agent | grok-4.6 | pass | pass | pass (gpt-6.1-sol) | pass | pass | Vercel AI Gateway | cache 66% of input || docker-agent | hunyuan-4-preview | pass | pass | pass (gpt-6.1-sol) | pass | pass | Vercel AI Gateway | cache 84% of input |@@ -6450,7 +6448,6 @@ Not run in this column, 595 pairs the provider serves that the harness did not r| fast-agent | gemini-3.6-flash | pass | pass | pass (gpt-6.1-sol) | pass | pass | Vercel AI Gateway | || fast-agent | gemini-3.7-flash | pass | pass | pass (gpt-6.1-sol) | pass | pass | Vercel AI Gateway | || fast-agent | gemini-3.8-flash | pass | pass | pass (gpt-6.1-sol) | pass | pass | Vercel AI Gateway | |-| fast-agent | gemini-4-argon | pass | pass | pass (gpt-6.1-sol) | pass | pass | Vercel AI Gateway | || fast-agent | glm-5.3 | pass | pass | pass (gpt-6.1-sol) | pass | pass | Vercel AI Gateway | || fast-agent | glm-5.3-flash | pass | pass | pass (gpt-6.1-sol) | pass | pass | Vercel AI Gateway | || fast-agent | gpt-5.2 | pass | pass | pass (gpt-6.1-sol) | pass | pass | Vercel AI Gateway | |@@ -6651,7 +6648,7 @@ Not run in this column, 595 pairs the provider serves that the harness did not r| vibe | qwen3.8-max | pass | pass | pass (gpt-6.1-sol) | pass | pass | Vercel AI Gateway | cache 17% of input || vibe | step-3.7-flash | pass | pass | pass (gpt-6.1-sol) | pass | pass | Vercel AI Gateway | cache 32% of input |-369 pairs, 1738 of 1756 scenario runs passed; 9 pairs served by another connection or as another model are findings, not counted.+367 pairs, 1728 of 1746 scenario runs passed; 9 pairs served by another connection or as another model are findings, not counted.Findings, pairs served by a connection other than the one under test or as a model other than the id asked for:@@ -6687,7 +6684,7 @@ Findings, pairs served by a connection other than the one under test or as a mod- qwen x claude-haiku-5.5: no prompt cache reads over 569929 input tokens (every call paid full price)- qwen x grok-4.1-fast: served as spacexai/grok-4.1-fast-reasoning (the CLI reports the model it ran)-Not run in this column, 1042 pairs the provider serves that the harness did not run, with the reason:+Not run in this column, 1044 pairs the provider serves that the harness did not run, with the reason:- agentzero x claude-fable-5: not run, not in this harness's catalog (unmeasured or excluded, see the catalog's note)- agentzero x claude-fable-5-1: not run, not in this harness's catalog (unmeasured or excluded, see the catalog's note)@@ -7049,6 +7046,7 @@ Not run in this column, 1042 pairs the provider serves that the harness did not- docker-agent x claude-opus-4.7: not run, not in this harness's catalog (unmeasured or excluded, see the catalog's note)- docker-agent x gpt-5.3-codex: not run, the model answers on the Responses API only and this harness speaks chat/completions only- docker-agent x gpt-6-astra: not run, the model answers on the Responses API only and this harness speaks chat/completions only+- docker-agent x grok-4.20: not run, not in this harness's catalog (unmeasured or excluded, see the catalog's note)- docker-agent x grok-4.5: not run, not in this harness's catalog (unmeasured or excluded, see the catalog's note)- docker-agent x grok-build-0.1: not run, not in this harness's catalog (unmeasured or excluded, see the catalog's note)- docker-agent x hunyuan-3: not run, not in this harness's catalog (unmeasured or excluded, see the catalog's note)@@ -7114,6 +7112,7 @@ Not run in this column, 1042 pairs the provider serves that the harness did not- dsh x qwen3.8-flash: not run, not run in this column- dsh x qwen3.8-max: not run, not run in this column- dsh x step-3.7-flash: not run, not run in this column+- fast-agent x gemini-4-argon: not run, not in this harness's catalog (unmeasured or excluded, see the catalog's note)- fast-agent x gpt-5.3-codex: not run, the model answers on the Responses API only and this harness speaks chat/completions only- fast-agent x gpt-6-astra: not run, the model answers on the Responses API only and this harness speaks chat/completions only- fast-agent x grok-4.5: not run, not in this harness's catalog (unmeasured or excluded, see the catalog's note)diff --git a/gateway/app.py b/gateway/app.pyindex f67f39c2..b8e92cc8 100644--- a/gateway/app.py+++ b/gateway/app.py@@ -8367,6 +8367,11 @@ _NOT_OFFERED: dict[str, frozenset[str]] = {# outside the workspace and never shown, 1 of 4 conversations, after# answering "hi" with some 40 tool calls (OpenRouter, 0.37.1 on the# test instance): what took it off dcode and fast-agent.+ # grok-4.20 asked to restyle a deck another model had made in the same+ # conversation (gpt-6.1-sol), said it had and made no tool call, on+ # the hosted service (2026-10-27). Off there and here alike (Richard:+ # "Hide those two pairs"), though the same two turns on the test+ # instance's route (Vercel, 0.37.7) made the change, 2 of 2.# fast-agent grok-4.5 answered DONE with no tool call, tools declared and history intact:# the matrix (Vercel) and 3 of 4 in sequence.# llama-3.3-70b answered the recycle check DONE, not its first word (OpenRouter).@@ -8388,6 +8393,12 @@ _NOT_OFFERED: dict[str, frozenset[str]] = {# instructions say is never shown (no file reached the console), 1 of 4# conversations; another ran about 150 shell commands for the same# file (OpenRouter, 0.37.1 on the test instance): what took it off dcode.+ # gemini-4-argon after another model's turn (gpt-6.1-sol, claude-haiku-4.5) whose+ # history holds an image attach_media added, on Google's own API: one+ # tool call and then an empty answer, or a promise and no tool call,+ # twice each (hosted, 2026-10-27). argon after argon, and the same+ # conversation without the image, pass, as does Vercel's route. Off+ # here as on the hosted service (Richard: "Hide those two pairs").# dcode grok-4.5 wrote its write_file call as JSON text, 1 of 4 in sequence.# kimi-k2.7-code answered the artifact turn DONE with no tool call (Vercel).# hunyuan-4-preview a follow-up streamed until Vercel ended it, about 13 minutes@@ -8409,9 +8420,9 @@ _NOT_OFFERED: dict[str, frozenset[str]] = {# rerun put its .pptx there (OpenRouter, 0.37.1 on the test instance)."vibe": frozenset({"gpt-5.6-luna", "llama-3.3-70b", "deepseek-v4.1-flash", "claude-opus-4.7"}),"docker-agent": frozenset({"grok-build-0.1", "grok-4.5", "claude-opus-4.7", "llama-3.3-70b",- "nemotron-3.5-lightning"}),+ "nemotron-3.5-lightning", "grok-4.20"}),"fast-agent": frozenset({"grok-4.5", "llama-3.3-70b", "kimi-k2.7-code", "hunyuan-4-preview", "qwen3.8-27b",- "qwen3.8-max", "nemotron-3-super", "nemotron-3.5-lightning"}),+ "qwen3.8-max", "nemotron-3-super", "nemotron-3.5-lightning", "gemini-4-argon"}),"dcode": frozenset({"grok-4.5", "kimi-k2.7-code", "hunyuan-4-preview", "muse-glimmer-30b", "muse-spark-1.1","llama-3.3-70b", "nemotron-3.5-lightning"}),}diff --git a/gateway/tests/test_gemini_4_argon.py b/gateway/tests/test_gemini_4_argon.pyindex 8cbca318..980468d6 100644--- a/gateway/tests/test_gemini_4_argon.py+++ b/gateway/tests/test_gemini_4_argon.py@@ -19,7 +19,16 @@ def test_each_vendor_that_serves_it_names_it_as_it_does():def test_every_base_that_offers_the_previous_pro_model_offers_it_first_in_its_family():for b, c in gw._MODEL_CATALOG.items():- if "gemini-3.1-pro-preview" in c["models"]:+ if "gemini-3.1-pro-preview" in c["models"] and "gemini-4-argon" not in gw._NOT_OFFERED.get(b, ()):gem = [m for m in c["models"] if m.startswith("gemini")]assert gem[0] == "gemini-4-argon", (b, gem)assert gw._MODEL_ORDER.count("gemini-4-argon") == 1+++def test_two_pairs_are_off_after_deck_conversations_that_changed_models():+ """fast-agent + gemini-4-argon and docker-agent + grok-4.20 failed on the hosted service in deck conversations+ that changed models (2026-10-27; Richard: "Hide those two pairs"). Each base keeps the rest of the family."""+ assert "gemini-4-argon" not in gw._MODEL_CATALOG["fast-agent"]["models"]+ assert "grok-4.20" not in gw._MODEL_CATALOG["docker-agent"]["models"]+ assert "gemini-3.8-flash" in gw._MODEL_CATALOG["fast-agent"]["models"]+ assert "grok-4.6" in gw._MODEL_CATALOG["docker-agent"]["models"]diff --git a/ui/src/lib/harness.ts b/ui/src/lib/harness.tsindex 77fbdc34..5d8f7c90 100644--- a/ui/src/lib/harness.ts+++ b/ui/src/lib/harness.ts@@ -261,12 +261,12 @@ export const OOB: OobHarness[] = [tools: [], skills: [] },{ id: 'docker-agent', name: 'Docker Agent', version: 'v1.152.0', backend: 'docker-agent', status: 'ready',// Placeholder only; the chat-completions list, for jcode's reason.- models: ['gpt-6.1-sol', 'gpt-6-sol', 'gpt-6-luna', 'gpt-5.6-sol', 'gpt-5.6-terra', 'gpt-5.6-luna', 'gpt-5.5', 'gpt-5.4', 'gpt-5.4-mini', 'gpt-5.2', 'claude-fable-5-1', 'claude-fable-5', 'claude-opus-5.5', 'claude-opus-5', 'claude-sonnet-5.5', 'claude-sonnet-5', 'claude-opus-4.8', 'claude-sonnet-4.6', 'claude-haiku-5.5', 'claude-haiku-4.5', 'gemini-4-argon', 'gemini-3.8-flash', 'gemini-3.7-flash', 'gemini-3.6-flash', 'gemini-3.5-flash', 'gemini-3.5-flash-lite', 'gemini-3.1-pro-preview', 'gemini-3.1-flash-lite', 'gemini-3-flash-preview', 'grok-4.6', 'grok-4.3', 'grok-4.20', 'muse-wonder-1.0', 'muse-spark-1.3', 'muse-spark-1.2', 'muse-spark-1.1', 'muse-glimmer-30b', 'deepseek-v4.1-flash', 'deepseek-v4-pro', 'deepseek-v4-flash', 'kimi-k3', 'kimi-k2.7-code', 'qwen3.8-max', 'qwen3.8-flash', 'qwen3.8-27b', 'qwen3.7-max', 'qwen3.7-plus', 'glm-5.3', 'glm-5.3-flash', 'mistral-medium-3.5', 'step-3.7-flash', 'hunyuan-4-preview', 'nemotron-3-super'], defaultModel: 'gpt-5.4', moreModels: 0,+ models: ['gpt-6.1-sol', 'gpt-6-sol', 'gpt-6-luna', 'gpt-5.6-sol', 'gpt-5.6-terra', 'gpt-5.6-luna', 'gpt-5.5', 'gpt-5.4', 'gpt-5.4-mini', 'gpt-5.2', 'claude-fable-5-1', 'claude-fable-5', 'claude-opus-5.5', 'claude-opus-5', 'claude-sonnet-5.5', 'claude-sonnet-5', 'claude-opus-4.8', 'claude-sonnet-4.6', 'claude-haiku-5.5', 'claude-haiku-4.5', 'gemini-4-argon', 'gemini-3.8-flash', 'gemini-3.7-flash', 'gemini-3.6-flash', 'gemini-3.5-flash', 'gemini-3.5-flash-lite', 'gemini-3.1-pro-preview', 'gemini-3.1-flash-lite', 'gemini-3-flash-preview', 'grok-4.6', 'grok-4.3', 'muse-wonder-1.0', 'muse-spark-1.3', 'muse-spark-1.2', 'muse-spark-1.1', 'muse-glimmer-30b', 'deepseek-v4.1-flash', 'deepseek-v4-pro', 'deepseek-v4-flash', 'kimi-k3', 'kimi-k2.7-code', 'qwen3.8-max', 'qwen3.8-flash', 'qwen3.8-27b', 'qwen3.7-max', 'qwen3.7-plus', 'glm-5.3', 'glm-5.3-flash', 'mistral-medium-3.5', 'step-3.7-flash', 'hunyuan-4-preview', 'nemotron-3-super'], defaultModel: 'gpt-5.4', moreModels: 0,systemPrompt: 'You are an autonomous coding agent working in a real git workspace with shell and file access, reading and editing files and running commands to complete the task end to end.',tools: [], skills: [] },{ id: 'fast-agent', name: 'fast-agent', version: 'v0.10.47', backend: 'fast-agent', status: 'ready',// Placeholder only; the chat-completions list, for jcode's reason.- models: ['gpt-6.1-sol', 'gpt-6-sol', 'gpt-6-luna', 'gpt-5.6-sol', 'gpt-5.6-terra', 'gpt-5.6-luna', 'gpt-5.5', 'gpt-5.4', 'gpt-5.4-mini', 'gpt-5.2', 'claude-fable-5-1', 'claude-fable-5', 'claude-opus-5.5', 'claude-opus-5', 'claude-sonnet-5.5', 'claude-sonnet-5', 'claude-opus-4.8', 'claude-opus-4.7', 'claude-sonnet-4.6', 'claude-haiku-5.5', 'claude-haiku-4.5', 'gemini-4-argon', 'gemini-3.8-flash', 'gemini-3.7-flash', 'gemini-3.6-flash', 'gemini-3.5-flash', 'gemini-3.5-flash-lite', 'gemini-3.1-pro-preview', 'gemini-3.1-flash-lite', 'gemini-3-flash-preview', 'grok-4.6', 'grok-4.3', 'grok-4.20', 'grok-build-0.1', 'muse-wonder-1.0', 'muse-spark-1.3', 'muse-spark-1.2', 'muse-spark-1.1', 'muse-glimmer-30b', 'deepseek-v4.1-flash', 'deepseek-v4-pro', 'deepseek-v4-flash', 'kimi-k3', 'qwen3.8-flash', 'qwen3.7-max', 'qwen3.7-plus', 'glm-5.3', 'glm-5.3-flash', 'mistral-medium-3.5', 'step-3.7-flash'], defaultModel: 'gpt-5.4', moreModels: 0,+ models: ['gpt-6.1-sol', 'gpt-6-sol', 'gpt-6-luna', 'gpt-5.6-sol', 'gpt-5.6-terra', 'gpt-5.6-luna', 'gpt-5.5', 'gpt-5.4', 'gpt-5.4-mini', 'gpt-5.2', 'claude-fable-5-1', 'claude-fable-5', 'claude-opus-5.5', 'claude-opus-5', 'claude-sonnet-5.5', 'claude-sonnet-5', 'claude-opus-4.8', 'claude-opus-4.7', 'claude-sonnet-4.6', 'claude-haiku-5.5', 'claude-haiku-4.5', 'gemini-3.8-flash', 'gemini-3.7-flash', 'gemini-3.6-flash', 'gemini-3.5-flash', 'gemini-3.5-flash-lite', 'gemini-3.1-pro-preview', 'gemini-3.1-flash-lite', 'gemini-3-flash-preview', 'grok-4.6', 'grok-4.3', 'grok-4.20', 'grok-build-0.1', 'muse-wonder-1.0', 'muse-spark-1.3', 'muse-spark-1.2', 'muse-spark-1.1', 'muse-glimmer-30b', 'deepseek-v4.1-flash', 'deepseek-v4-pro', 'deepseek-v4-flash', 'kimi-k3', 'qwen3.8-flash', 'qwen3.7-max', 'qwen3.7-plus', 'glm-5.3', 'glm-5.3-flash', 'mistral-medium-3.5', 'step-3.7-flash'], defaultModel: 'gpt-5.4', moreModels: 0,systemPrompt: 'You are an autonomous coding agent. You work on a real git workspace with shell and file access, reading and editing files and running commands to complete the task end to end.',tools: [], skills: [] },{ id: 'dcode', name: 'Deep Agents Code', version: 'v0.2.1', backend: 'dcode', status: 'ready',