diff --git a/CHANGES b/CHANGES
index 3abb8a1..bf716b1 100644
--- a/CHANGES
+++ b/CHANGES
@@ -1,5 +1,9 @@
 # Changelog
 
+## v1.6.1 — 2026-08-21
+- REPL: show time to task completion when each model finishes, persist timings in the session, and include last/total duration in `/report`
+- `agent-tester run` reports: add a fastest-first time-to-completion ranking among successful agents
+
 ## v1.6.0 — 2026-08-21
 - Add `type: cursor` REPL provider (Cursor CLI) so Auto / Router works in `agent-tester repl` alongside other models; `/reset` clears the CLI chat session; document `models.cursor-auto` config
 
diff --git a/README.md b/README.md
index 091a0b7..19d0729 100644
--- a/README.md
+++ b/README.md
@@ -367,7 +367,7 @@ The REPL fans out each prompt to all configured models in parallel and maintains
 | `/status` | Show which models are running or idle |
 | `/stop [@model …]` | Cancel a running model. Without a tag, stops all busy models. |
 | `/interrupt [@model …] <message>` | Cancel and immediately re-dispatch with `<message>`. Without a tag, targets all busy models. |
-| `/report` | Show each model's git commits, diff stats, and token usage. Report data is persisted in the session file (`~/.config/agenttester/sessions/<name>.yaml`). |
+| `/report` | Show each model's git commits, diff stats, token usage, and time to completion. Report data is persisted in the session file (`~/.config/agenttester/sessions/<name>.yaml`). |
 | `/evaluate [m1,m2,…]` | Cross-evaluate: each model reviews the others' work. Evaluation documents are saved to `.agenttester/evaluations/<session>/`. Eval results are also persisted in the session file. |
 | `/iterate <prompt>` | After `/evaluate`, inject peer evaluations as context and send an iteration prompt. Requires `y` confirmation. |
 
diff --git a/pyproject.toml b/pyproject.toml
index 1c3a493..a519f07 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -4,7 +4,7 @@ build-backend = "hatchling.build"
 
 [project]
 name = "agenttester"
-version = "1.6.0"
+version = "1.6.1"
 description = "Run a prompt against multiple coding agents in parallel and compare results"
 readme = "README.md"
 requires-python = ">=3.10"
diff --git a/src/agenttester/repl.py b/src/agenttester/repl.py
index a3d7e7a..8b6db22 100644
--- a/src/agenttester/repl.py
+++ b/src/agenttester/repl.py
@@ -46,6 +46,18 @@
 _CTRL_C_TIMEOUT = 2.0
 _BRANCH_SLUG_MAX_LEN = 60
 
+
+def _format_duration(seconds: float) -> str:
+    """Human-readable duration for task completion display."""
+    if seconds < 60:
+        return f"{seconds:.1f}s"
+    minutes, secs = divmod(seconds, 60)
+    if minutes < 60:
+        return f"{int(minutes)}m {secs:.0f}s"
+    hours, minutes = divmod(int(minutes), 60)
+    return f"{hours}h {minutes}m"
+
+
 _SLASH_COMMANDS = [
     ("/reset", "clear conversation history"),
     ("/status", "show running/idle models"),
@@ -327,12 +339,25 @@ async def _run_one(
     model: Model,
     prompt: str,
     on_event: Callable[[str, str], None] | None = None,
-) -> tuple[str, str]:
+) -> tuple[str, str, float]:
+    """Run one model query. Returns ``(name, reply, duration_seconds)``."""
+    start = time.monotonic()
     try:
         r = await _query_async(model, prompt, on_event=on_event)
     except Exception as exc:
         r = str(exc)
-    return name, r
+    return name, r, time.monotonic() - start
+
+
+def _print_model_done(console: Console, name: str, reply: str, duration: float) -> None:
+    """Print the per-model completion line including time to task completion."""
+    elapsed = _format_duration(duration)
+    if reply.startswith("[error]"):
+        console.print(f"  [red]✗ {name}[/red]: {reply[:100]}  [dim]({elapsed})[/dim]")
+    else:
+        console.print(
+            f"  [green]✓[/green] [bold]{name}[/bold]: done  [dim]({elapsed})[/dim]"
+        )
 
 
 def _clean_name(raw: str) -> str:
@@ -460,6 +485,7 @@ async def _run_report(
     console: Console,
     reports_store: dict[str, dict[str, str]],
     token_usage: dict[str, dict[str, dict[str, int]]] | None = None,
+    query_timings: dict[str, dict[str, float | int]] | None = None,
 ) -> None:
     """Collect each model's work and display a summary. Populates *reports_store*."""
 
@@ -478,6 +504,22 @@ async def _fetch(name: str, model: Model) -> tuple[str, dict[str, str]]:
         console.print(
             "[yellow]No models have committed or uncommitted work yet.[/yellow]\n"
         )
+        # Still show timings even when there is no git work yet.
+        if query_timings:
+            console.print("[bold]Time to completion[/bold]")
+            for name in models:
+                timing = query_timings.get(name)
+                if not timing:
+                    continue
+                last = float(timing.get("last", 0))
+                total = float(timing.get("total", 0))
+                count = int(timing.get("count", 0))
+                console.print(
+                    f"  [bold]{name}[/bold]: last {_format_duration(last)}"
+                    f" · total {_format_duration(total)} ({count} quer"
+                    f"{'y' if count == 1 else 'ies'})"
+                )
+            console.print()
         return
 
     console.print(f"\n[bold]Work report — {len(models_with_work)} model(s)[/bold]\n")
@@ -488,6 +530,16 @@ async def _fetch(name: str, model: Model) -> tuple[str, dict[str, str]]:
                 console.print(f"  [dim]{line}[/dim]")
         if report["stat"]:
             console.print(f"  {report['stat']}")
+        if query_timings and name in query_timings:
+            timing = query_timings[name]
+            last = float(timing.get("last", 0))
+            total = float(timing.get("total", 0))
+            count = int(timing.get("count", 0))
+            console.print(
+                f"  [dim]time: last {_format_duration(last)}"
+                f" · total {_format_duration(total)} ({count} quer"
+                f"{'y' if count == 1 else 'ies'})[/dim]"
+            )
         if token_usage and name in token_usage:
             for phase, counts in token_usage[name].items():
                 in_t = counts.get("input", 0)
@@ -1003,7 +1055,7 @@ async def _iterate_run(
                                 _in_before = _m.input_tokens
                                 _out_before = _m.output_tokens
                                 try:
-                                    _, reply = await _run_one(
+                                    _, reply, duration = await _run_one(
                                         _nm,
                                         _m,
                                         _p,
@@ -1021,19 +1073,15 @@ async def _iterate_run(
                                     session.add_tokens(
                                         _nm, "queries", delta_in, delta_out
                                     )
+                                session.add_timing(_nm, duration)
                                 if _m.event_logger is not None:
                                     _m.event_logger.log("response", reply)
                                     _m.event_logger.log(
-                                        "status", "waiting for next instructions"
-                                    )
-                                if reply.startswith("[error]"):
-                                    console.print(
-                                        f"  [red]✗ {_nm}[/red]: {reply[:100]}"
-                                    )
-                                else:
-                                    console.print(
-                                        f"  [green]✓[/green] [bold]{_nm}[/bold]: done"
+                                        "status",
+                                        f"waiting for next instructions"
+                                        f" (completed in {_format_duration(duration)})",
                                     )
+                                _print_model_done(console, _nm, reply, duration)
 
                             _had_user_input = True
                             _it2 = asyncio.create_task(_iterate_run())
@@ -1130,7 +1178,11 @@ async def _iterate_run(
 
                 async def _report_task() -> None:
                     await _run_report(
-                        models, console, _reports, token_usage=session.token_usage
+                        models,
+                        console,
+                        _reports,
+                        token_usage=session.token_usage,
+                        query_timings=session.query_timings,
                     )
                     session.reports = dict(_reports)
                     session.save()
@@ -1331,7 +1383,7 @@ async def _background_run(
                     _in_before = _m.input_tokens
                     _out_before = _m.output_tokens
                     try:
-                        _, reply = await _run_one(
+                        _, reply, duration = await _run_one(
                             _nm,
                             _m,
                             _prompt,
@@ -1347,13 +1399,15 @@ async def _background_run(
                     delta_out = _m.output_tokens - _out_before
                     if delta_in or delta_out:
                         session.add_tokens(_nm, "queries", delta_in, delta_out)
+                    session.add_timing(_nm, duration)
                     if _m.event_logger is not None:
                         _m.event_logger.log("response", reply)
-                        _m.event_logger.log("status", "waiting for next instructions")
-                    if reply.startswith("[error]"):
-                        console.print(f"  [red]✗ {_nm}[/red]: {reply[:100]}")
-                    else:
-                        console.print(f"  [green]✓[/green] [bold]{_nm}[/bold]: done")
+                        _m.event_logger.log(
+                            "status",
+                            f"waiting for next instructions"
+                            f" (completed in {_format_duration(duration)})",
+                        )
+                    _print_model_done(console, _nm, reply, duration)
 
                 _had_user_input = True
                 _bt = asyncio.create_task(_background_run())
diff --git a/src/agenttester/report.py b/src/agenttester/report.py
index 939328e..76e6451 100644
--- a/src/agenttester/report.py
+++ b/src/agenttester/report.py
@@ -51,6 +51,20 @@ def generate_report(
             f"| -{stats.deletions} |"
         )
 
+    # Rank by time to task completion (fastest first among successful agents)
+    finished = sorted(
+        (r for r in results if r.exit_code == 0 and not r.error),
+        key=lambda r: r.duration,
+    )
+    if len(finished) >= 2:
+        lines.extend(
+            [
+                "",
+                "**Time to completion** (fastest first): "
+                + ", ".join(f"{r.agent_name} ({r.duration:.1f}s)" for r in finished),
+            ]
+        )
+
     lines.append("")
 
     # Per-agent details
diff --git a/src/agenttester/session.py b/src/agenttester/session.py
index b1d93ac..cf68f0b 100644
--- a/src/agenttester/session.py
+++ b/src/agenttester/session.py
@@ -42,6 +42,8 @@ class ReplSession:
     reports: dict[str, dict[str, str]] = field(default_factory=dict)
     eval_results: dict[str, dict[str, str]] = field(default_factory=dict)
     token_usage: dict[str, dict[str, dict[str, int]]] = field(default_factory=dict)
+    # model -> {total: float seconds, last: float, count: int}
+    query_timings: dict[str, dict[str, float | int]] = field(default_factory=dict)
 
     @classmethod
     def create(cls, name: str) -> ReplSession:
@@ -69,6 +71,7 @@ def load(cls, name: str, sessions_dir: Path | None = None) -> ReplSession:
             reports=data.get("reports", {}),
             eval_results=data.get("eval_results", {}),
             token_usage=data.get("token_usage", {}),
+            query_timings=data.get("query_timings", {}),
         )
 
     @classmethod
@@ -94,6 +97,15 @@ def add_tokens(
         phase_usage["input"] += in_tok
         phase_usage["output"] += out_tok
 
+    def add_timing(self, model_name: str, seconds: float) -> None:
+        """Record a query duration for *model_name* (last + cumulative)."""
+        timing = self.query_timings.setdefault(
+            model_name, {"total": 0.0, "last": 0.0, "count": 0}
+        )
+        timing["last"] = float(seconds)
+        timing["total"] = float(timing.get("total", 0.0)) + float(seconds)
+        timing["count"] = int(timing.get("count", 0)) + 1
+
     def save(
         self, sessions_dir: Path | None = None, max_sessions: int | None = None
     ) -> None:
@@ -110,6 +122,7 @@ def save(
                     "reports": self.reports,
                     "eval_results": self.eval_results,
                     "token_usage": self.token_usage,
+                    "query_timings": self.query_timings,
                 },
                 default_flow_style=False,
                 allow_unicode=True,
diff --git a/tests/test_repl.py b/tests/test_repl.py
index d99ddba..6621e3f 100644
--- a/tests/test_repl.py
+++ b/tests/test_repl.py
@@ -337,17 +337,31 @@ async def test_returns_name_and_result(self) -> None:
         provider = MagicMock()
         provider.async_call = AsyncMock(return_value="ok")
         model = Model(name="llama3", model_id="llama", provider=provider)
-        name, result = await _run_one("llama3", model, "hello")
+        name, result, duration = await _run_one("llama3", model, "hello")
         assert name == "llama3"
         assert result == "ok"
+        assert duration >= 0
 
     async def test_returns_error_string_on_exception(self) -> None:
         provider = MagicMock()
         provider.async_call = AsyncMock(side_effect=OSError("unreachable"))
         model = Model(name="llama3", model_id="llama", provider=provider)
-        name, result = await _run_one("llama3", model, "hello")
+        name, result, duration = await _run_one("llama3", model, "hello")
         assert name == "llama3"
         assert "[error]" in result
+        assert duration >= 0
+
+
+class TestFormatDuration:
+    def test_seconds(self) -> None:
+        from agenttester.repl import _format_duration
+
+        assert _format_duration(12.34) == "12.3s"
+
+    def test_minutes(self) -> None:
+        from agenttester.repl import _format_duration
+
+        assert _format_duration(125) == "2m 5s"
 
 
 # ---------------------------------------------------------------------------
diff --git a/tests/test_report.py b/tests/test_report.py
index 125e7d0..e0fc1be 100644
--- a/tests/test_report.py
+++ b/tests/test_report.py
@@ -53,6 +53,19 @@ def test_contains_summary_table(self) -> None:
         assert "✅" in report
         assert "❌" in report
 
+    def test_time_to_completion_ranking(self) -> None:
+        results = [
+            AgentResult("slow", 0, 30.0, "", "", None),
+            AgentResult("fast", 0, 5.0, "", "", None),
+            AgentResult("failed", 1, 1.0, "", "", "boom"),
+        ]
+        report = generate_report("r-time", "b" * 40, "test", results, _mock_git())
+        assert "**Time to completion**" in report
+        # Fastest successful agent listed first
+        ranking = report.split("**Time to completion**")[1].split("\n")[0]
+        assert ranking.index("fast") < ranking.index("slow")
+        assert "failed" not in ranking
+
     def test_per_agent_sections(self) -> None:
         results = [
             AgentResult("agent1", 0, 3.0, "", "", None),
diff --git a/tests/test_session.py b/tests/test_session.py
index 684b093..e06f094 100644
--- a/tests/test_session.py
+++ b/tests/test_session.py
@@ -136,3 +136,21 @@ def test_deletes_file(self, tmp_path: Path) -> None:
     def test_delete_missing_is_noop(self, tmp_path: Path) -> None:
         s = ReplSession.create("ghost")
         s.delete(tmp_path)  # should not raise
+
+
+class TestQueryTimings:
+    def test_add_timing_accumulates(self) -> None:
+        s = ReplSession.create("t")
+        s.add_timing("m1", 1.5)
+        s.add_timing("m1", 2.5)
+        assert s.query_timings["m1"]["last"] == 2.5
+        assert s.query_timings["m1"]["total"] == 4.0
+        assert s.query_timings["m1"]["count"] == 2
+
+    def test_timings_roundtrip(self, tmp_path: Path) -> None:
+        s = ReplSession.create("timed")
+        s.add_timing("cursor-auto", 12.3)
+        s.save(tmp_path)
+        loaded = ReplSession.load("timed", tmp_path)
+        assert loaded.query_timings["cursor-auto"]["last"] == 12.3
+        assert loaded.query_timings["cursor-auto"]["count"] == 1